Book Tax Diff and Earning Growth

BOOK-TAX DIFFERENCES AND EARNINGS GROWTH
by
MARK JACKSON
A DISSERTATION
Presented to the Department of Accounting
and the Graduate School of the University of Oregon
in partial fulfillment of the requirements
for the degree of
Doctor of Philosophy
June 2009
11
University of Oregon Graduate School

Confirmation of Approval and Acceptance of Dissertation prepared by:
Mark Jackson
Title:
"Book-Tax Differences and Earnings Growth"
This dissertation has been accepted and approved in partial fulfillment of the requirements for
the Doctor of Philosophy degree in the Department of Accounting by:
David Guenther, Chairperson, Accounting
Steven Matsunaga, Member, Accounting
Linda Krull, Member, Accounting
Glen Waddell, Outside Member, Economics
and Richard Linton, Vice President for Research and Graduate Studies/Dean of the Graduate
School for the University of Oregon.
June 13,2009
Original approval signatures are on file with the Graduate School and the University of Oregon
Libraries.
111
An Abstract of the Dissertation of

Mark Jackson
in the Depmiment of Accounting
for the degree of

to be taken
Doctor of Philosophy
June 2009
Title: BOOK-TAX DIFFERENCES AND EARNINGS GROWTH
Approved:
Dr. David Guenther
I examine the relation between book-tax differences (BTDs) and earnings growth.
Because financial accounting rules afford managers more flexibility and discretion in
repOliing than tax accounting rules, prior studies suggest that large differences between
book and taxable income indicate lower quality (or less persistent) earnings. Lev and
Nissim and Hanlon provide evidence that BTDs contain infonnation about future finn
performance, but the nature of the causality in this relation is not clear. While BTDs
could proxy for earnings quality, they may also reveal underlying economic events or
management's private information about future perfonnance or simply predict future
reversals in effective tax rates.
I divide total BTDs into their measurable components: temporary (defelTed taxes)
and non-temporary (pennanent differences and tax accruals), and test their relation with
the components of net income changes: pretax earnings changes and tax expense
changes. I hypothesize that the non-temporary component ofBTDs is negatively related
IV
to future changes in tax expense, whereas the temporary component ofBTDs is

negatively related to changes in future pretax eall1ings. I also examine the maintained
hypothesis that the lower earnings growth for large BTD finTIs is due to earnings
management. I use various proxies from prior literature to identify fin11s potentially
managing earnings and test whether the presence or absence of suspected earnings
management activity alters the relation between BTDs and earnings changes.
My results provide compelling evidence that pell11anent BTDs are related only to
future changes in tax expense, and temporary BTDs are related to changes in pretax
eall1ings. These results are robust to multiple sensitivity analyses, including a replication
of the sample and methodology of Lev and Nissim. The results also hold in the case of
fill11s not suspected of eall1ings management. In fact, I find only limited evidence that the
results are stronger in the presence of earnings management. Overall, my study suggests
that it is only the temporary component of BTDs that is related to future firm
performance, with non-temporary differences being related to future tax expense changes,
and that these results are primarily due to underlying economic factors, not eall1ings
management.
CURRICULUM VITAE
NAME OF AUTHOR: Mark Jackson
PLACE OF BIRTH: Torrance, California
DATE OF BIRTH: December 29,1963
GRADUATE AND UNDERGRADUATE SCHOOLS ATTENDED:

University of Oregon, Eugene
University of Texas, El Paso
DEGREES AWARDED:
Doctor of Philosophy, Accounting, June 2009, University of Oregon
Bachelor of Science, Accounting, May 2005, University of Texas, El Paso
AREAS OF SPECIAL INTEREST:

The interaction of financial and tax accounting systems
Earnings management and firm value
PROFESSIONAL EXPERIENCE:
Graduate teaching fellow, University of Oregon, 2005-present
Owner/Manager, Jackson Bookkeeping and Tax Preparation, 1998-present
GRANTS, AWARDS AND HONORS:

Accounting Circle at Lundquist College of Business Summer Fellowship,
2007
Beta Gamma Sigma, business honor fraternity, admitted 2005
Summa Cum Laude, University of Texas, El Paso, 2005
VI
ACKNOWLEDGMENTS
I am thankful for the guidance of my dissertation committee: Dave Guenther
(chair), Linda Krull, Steve Matsunaga, and Glen Waddell. This paper has also benefited
from discussions with Lisa Eiler, Josh Filzen, Isho Tama-Sweet, Dave Weber, and
workshop participants at Central Missouri University, Eastern Michigan University,
Louisiana State University, Texas Tech University, University of Oregon, University of
Nevada, Reno, and University ofNorth Texas.
VB
TABLE OF CONTENTS
Chapter
Page
I. INTRODUCTION
II. RELATED RESEARCH.........................................................................................
Temporary Differences
Total Differences
III. HYPOTHESIS DEVELOPMENT.........................................................................
11
Tax Accruals..........................................................................................................
13
Permanent Differences
16
Telnporary Differences
17
IV. RESEARCH DESIGN...........................................................................................
22
V. SAMPLE AND SUMMARY STATISTICS
30
Sample Selection....................................................................................................
30
Summary Statistics...............
31
VI. EMPIRICAL RESULTS

The Association between Book-Tax Differences and Earnings Changes
34
34
Impact of Earnings Management on Association between Book-Tax

Differences and Earnings Growth....................................................................
37
VII. ROBUSTNESS AND SENSITIVITY TESTS.....................................................
45
Reconciliation with Lev and Nissim (2004)
45
V111
Chapter
Page
Firm Fixed Effects
51
Negative Income and Tax Expense........................................................................
53
Ranked Regressions and Measurement of Cash from Operations
53
Absolute Values
55
Foreign Operations.................................................................................................
58
Earnings Managelnent
58
VIII. CONCLUSION
61
BIBLIOGRAPHy.....
63
IX
LIST OF FIGURES
Figure
1. Five-Year Growth in Net Income and its Components for Portfolios of Firms
Sorted by BTD and Cash Flow Measures........................................................
Page
47
LIST OF TABLES
Table
Page
1. Example of Relation between Book-Tax Difference

and Future Performance...................................................................................
14
2. Means of Effective Tax Rates (ETR) Across Ranks ofBTD

For Five Years Prior to and After BTD Ranking.............................................
15
3. Comparison of Variable Means Across Extremes in BTDs
32
4. Correlation Matrix
33
5. Regressions of Future Earnings Changes on BTD Measures and Other

Indicators of Growth
35
6. Examining Effect of EM on BTD Coefficients: Small Earnings Changes............
39
7. Examining Effect of EM on BTD Coefficients: Narrowly Avoiding a Loss ........
40
8. Examining Effect of EM on BTD Coefficients: Analysts' Forecasts....................
42
9. Examining Effect of EM on BTD Coefficients: Discretionary Accruals
43

Indicators of Growth: Lev and Nissim (2004) Methodology................................
49

Indicators of Growth: Firm Fixed Effects..............................................................
52

Indicators of Growth: Including Firms with Negative Income and Tax................
54

Indicators of Growth: Ranked Regressions
56

Indicators of Growth: Using Absolute Values of BTDs
57
15. Examining Effect of Foreign Operations on BTD Coefficients
59
CHAPTER I
INTRODUCTION
In this paper I examine the relation between book-tax differences (BTDs) and
earnings growth. Prior literature (Lev and Nissim, 2004; Hanlon, 2005) provides
evidence that BTDs contain information about future firm performance, but the nature of
the causality in this relation is not clear. Lev and Nissim (2004) suggest that BTDs
capture earnings management activity, or that tax accounting better captures 'core'
earnings. As tax accounting is more closely associated with current cash flows, the
magnitude of BTDs could capture the extent to which book income ventures away from
its 'permanent' levels. However, to the extent that BTDs capture underlying economic
items that are transitory in nature (such as a goodwill write-off or restructuring charge),
any relation between BTDs and earnings growth could simply be related to these events.
It is also possible that the book-tax difference predicts future variation in effective tax
rates (and therefore tax expense).
To examine these issues, I study the effect of the temporary and permanent
components of BTDs on future changes in both pretax earnings and tax expense. If BTDs
(or one of its components) contains infonnation about the future economic performance
of the firm, I expect to find a relation between this measure and future changes in pretax
earnings. However, ifBTDs (or its components) contains information about future
effective tax rates, then I expect to find a relation between this measure and future
changes in tax expense. In addition, prior literature is unsettled as to which measure of
BTD better proxies for earnings quality, as Lev and Nissim (2004) and Hanlon (2005)
find conflicting results using different BTD measures. By examining the effect of
different components ofBTDs on earnings growth, my research design reconciles the
conflicting findings of previous research, and provides guidance for future work
examining the relation between BTDs and future firm performance. Finally, I examine
the relation between BTDs and earnings growth in the presence and absence of earnings
management. If the BTD/earnings growth relation holds even in the absence of earnings
management, this would suggest that any relation between BTDs and earnings growth is
due at least in part to mechanical and economic factors rather than opportunistic
behavior. Understanding the relation between BTDs and future earnings changes is
important because it provides evidence on the usefulness of taxable income in
determining firm value.
Book-tax differences arise when different accounting systems are applied to the
same set of underlying economic events. Recent research has attempted to detennine
whether these differences are informative about a firm's earnings characteristics and
future performance. Lev and Nissim (2004) suggest that BTDs are related to growth
because they reflect earnings management activities that are not persistent or capture the
extent to which book earnings deviate from their permanent level. They predict and find
that BTDs are positively related to future earnings growth. 1 However, because they
I It should be noted that although Lev and Nissim (2004) use the tax to book ratio to capture the
impact of total BTDs on earnings growth, their inclusion of the temporary component, deferred taxes,
in their tests implies that the coefficient on the tax to book ratio is really capturing the impact of the
permanent component of BTDs on earnings growth. I discuss this issue and my approach in the
research design section.
3
measure growth as changes in net income, the growth they document could be related to
changes in underlying pretax earnings, or to changes in future income tax expense. To
distinguish these two effects, I separate growth in net income into its two components,
pretax earnings and tax expense, and separate total BTDs into its two components,
permanent and temporary differences. I then examine the effect of the two components of
BTDs on the growth of each income component. I find that permanent BTDs are not
positively related to growth in pretax earnings, but negatively related to changes in tax
expense, suggesting that permanent BTDs are more relevant in predicting future effective
tax rates than future core earnings growth.
Lev and Nissim (2004) also examine the relation between temporary BTDs
(deferred taxes) and earnings growth, and find no evidence of a relation. In contrast,
Hanlon (2005) finds evidence of lower earnings persistence for firms with large
temporary BTDs. In an effort to reconcile these conflicting results, I examine the relation
between temporary BTDs and the two components of earnings growth, pretax earnings
and tax expense. Consistent with Hanlon (2005), I find that growth in pretax income is
negatively related to temporary BTDs. My findings, that changes in future tax expense is
negatively related to non-temporary BTDs (permanent differences and tax accruals),
while growth in pretax income is negatively related to temporary BTDs, provides
evidence about which BTD measure is most appropriate in a given research setting, such
as investigations into how market participants use the BTD information in their decision
processes. 2
mak mg
A common theme in the BTD literature, and in the public press, is the concept
that BTDs proxy for earnings quality or represents earnings management. For example,
Rep. Lloyd Doggett (D-Tex.) referred to "a corporate culture of creative accounting and
reporting abuses" when he introduced legislation requiring companies to disclose and
explain the gap between book and taxable income (Weisman 2002, AOl). Phillips,
Pincus, and Rego (2003) present evidence that suggests that large BTDs are associated
with various measures of earnings management. Hanlon and Krishnan (2006) interpret
evidence of higher audit fees for firms with large BTDs as evidence that auditors
associate large BTDs with increased risk of earnings management. On the other hand,
Tang (2007) and Dhaliwal, Huber, Lee, and Pincus (2008) argue that much of the BTD is
due to mechanical or economic differences, unrelated to opportunistic behavior. To better
understand the relation between earnings growth and BTDs, I control for various proxies
of earnings management. While the relation between growth and BTDs holds for films
with no indication of opportunistic behavior, I find only weak evidence that firm years
suspected of earnings management have a stronger growth/BTD relation, suggesting that
it is principally underlying economic events, and not earnings management, that is
responsible for the relation between earnings growth and book-tax differences.
Prior literature is unsettled as to the appropriate measure for BTDs. Phillips, Pincus, and Rego
(2003) and Joos, Pratt, and Young (2000) focus on temporary differences, Weber (2008) and
Dhaliwal, Huber, Lee, and Pincus (2008) focus on total BTDs, while Hanlon and Krishnan (2006) use
both. See the following chapter for a discussion of these papers.
2
5
This research makes several contributions. First, while plior literature suggests
that BTDs are related to future earnings growth, it is not clear whether this growth is
related to future changes in core economic performance, future changes in tax expense, or
both. I provide evidence that growth in the two components of net income, pretax
earnings and tax expense, are related in different ways to BTDs. Second, I address the
unsettled question as to which BTD measure best predicts earnings growth. I find that
temporary differences predict growth in pretax earnings, while permanent differences
predict earnings growth related to changes in tax expense. These findings also help
reconcile the conflicting results of Lev and Nissim (2004) and Hanlon (2005) on the
relation between temporary BTDs and earnings growth/persistence. Third, this study
answers the call of Graham, Raedy, and Shackelford (2008) and Hanlon and Heitzman
(2009) for an examination ofthe components ofBTDs. Although Graham, Raedy, and
Shackelford (2008) ultimately desire a study that examines the specific accounts that
leads to BTDs and why they are informative, my study is a first step in at least breaking
BTDs into their temporary and permanent components, and understanding why and how
each component is informative. Hanlon and Heitzman (2009) specifically ask for an
examination of Lev and Nissim's (2004) results on non-temporary BTDs, indicating that
a better understanding of their results would progress the literature. Finally, I contribute
to the debate regarding book-tax conformity. Proponents of bridging the book-tax gap
assert that the gap exists due to opportunistic behavior, but my results suggest that the
relation between earnings growth and BTDs are more the result of underlying economic
events that are manifest in book-tax differences.
The remainder of the paper is organized as follows. The next Chapter highlights
related literature. Chapter III develops the hypotheses and Chapter IV discusses the
research design. Chapter V describes the sample employed in the empirical tests and
Chapter VI presents the results of those tests. Chapter VII presents robustness checks. I
provide concluding comments and avenues for future research in Chapter VIII.
CHAPTER II
RELATED RESEARCH
Recent research examines the association between BTDs and earnings quality.
Mills and Newberry (2001) find evidence consistent with firms increasing BTDs when
the nontax costs of conforming book to tax income outweigh the tax-related costs of nonconformity. Public firms, those with high debt or facing financial distress, as well as
those near bonus plan thresholds or with specific earnings patterns were associated with
larger BTDs. Their measure of BTDs, the difference between pretax book income and
taxable income, is based on firm level tax return data, information generally not available
to researchers or investors. In an effort to measure BTDs using publicly available
financial statements, two proxies have emerged: temporary differences, based on deferred
tax expense, and total differences, computed as the difference between book income and
grossed up current tax expense (taxable income). Both measures have been used to
measure earnings quality in the presence ofBTDs.
Using deferred taxes to represent temporary BTDs, Phillips, Pincus, and Rego
(2003) find that BTDs are incrementally useful beyond accruals in detecting some types
of earnings management. Joos, Pratt, and Young (2000) also examine temporary
differences and find that large BTD finns have weaker earnings to returns relations. They
interpret their results as suggesting that large temporary BTDs proxy for earnings
management, and that investors react to this proxy by not putting as much weight on
earnings in valuation. Hanlon (2005) finds evidence that large temporary BTDs (deferred
taxes) are informative about earnings persistence. Specifically, she finds that firms with
both large positive and large negative BTDs have less persistent earnings. She also finds
evidence that investors interpret large BTDs as a 'red flag,' and reduce their expectation
of future earnings persistence for these firm years.
Total D(fferences
Lev and Nissim (2004) develop a "tax-based fundamental" defined as the ratio of
estimated net taxable income to net book income. This ratio captures all book-tax
differences, both temporary and permanent, along with tax accruals. They hypothesize
and find that higher tax to book ratios are associated with higher levels of future earnings
growth. In contrast to the results of Hanlon (2005), they find that deferred taxes, the
temporary component ofBTDs, are not incrementally useful in predicting earnings
growth. Following Lev and Nissim (2004), Weber (2008) also uses total differences
(permanent, temporary, and tax accruals) when measuring BTDs, and finds that analysts
do not react efficiently to the information in this measure. Dhaliwal, Huber, Lee, and
Pincus (2008) also use total BTDs when examining the variability ofBTDs, and find that
the temporal variation ofBTDs is positively related to a firm's cost of capital. In
addition, they separate BTD variability into its economic and unexplained components,
and find that each is positively associated with cost of capital, suggesting that BTD
variability reflects information about both the firm's underlying economic volatility and
earnings management activity.
Underscoring the uncertainty in the literature as to which BTD measure is the

better proxy for earnings quality, Hanlon and Krishnan (2006) use both temporary
differences (deferred taxes) and total differences when testing if auditors use information
reflected in BTDs when setting audit fees. After controlling for other predictors of audit
fees, they find large BTDs associated with higher audit fees, consistent with BTDs
reflecting information about earnings quality and auditors' assessment of the risk
associated with auditing such statements. They find this result for both temporary and
total BTDs, but because they do not test the two together, it is not clear which (or both)
contain incremental information about earnings quality.
Total book tax differences have three components: temporary differences,
permanent differences, and tax accruals. As indicated above, temporary differences have
been investigated for their impact on earnings persistence and in identifying earnings
management. Pemlanent differences have been examined for their link with abusive tax
shelters. For example, Shevlin (2002) discusses the 'ideal' tax shelter as one that reduces
taxable income but never reduces book income, leading to permanent differences.
However, there has been little research attempting to link permanent BTDs to the quality
or growth of book income.
There is growing evidence on the use of the tax accruals component of total BTDs
in earnings management. Miller and Skinner (1998), Visvanathan (1998), and Bauman,
Bauman, and Halsey (2001) fail to find evidence of earnings management using the
valuation allowance account (VAA), but more recently Schrand and Wong (2003) and
Frank and Rego (2006) find evidence associating the VAA with earnings management.
10
Two concurrent studies by Blouin and Tuna (2007) and Gupta and Laux (2008) find
evidence of managing the tax contingency accrual (tax cushion). Krull (2004) finds
evidence that large international firms use the permanently reinvested earnings (PRE)
designation to manage their earnings. In a test for earnings management on aggregate tax
accruals, Dhaliwal, Gleason, and Mills (2004) find evidence that suggests firms adjust
their effective tax rate from the 3rd to the 4th quarter in order to meet earnings targets.
While the extant literature suggests that tax accruals are used to manage earnings,
it should be noted that any impact on earnings from manipulating tax accruals is
accomplished via changes in tax expense, not changes in pretax earnings. While the
literature suggests that either temporary or total BTDs can proxy for earnings quality, it is
not clear how the tax accruals components of total BTDs provide incrementally useful
information beyond that already contained in the temporary component in predicting
growth in pretax income. 3 It thus is an empirical question as to which BTD measure
better predicts earnings growth, and which component of growth (pretax income or
changes in tax expense) is related to the BTD. Additionally, most of the BTD literature
noted above operates under the maintained hypothesis that any BTD/earnings growth
relation is due to earnings management. However, because BTDs can also arise due to
underlying economic factors, it is an empirical question as to whether it is earnings
management or underlying economic events that drive the relation between BTDs and
earnings growth. I take up these issues in the following chapters.
3 Changes in the valuation allowance account can predict future changes in pretax earnings, as
changes in this account reflect management's perception of the finn's expected future perfonnance.
However, changes in the valuation allowance are captured as a deferred tax, and thus empirically will
be a temporary difference.
11
CHAPTER III
HYPOTHESIS DEVELOPMENT
Lev and Nissim (2004) document a relation between total BTDs and growth in net
income. Although their results are robust to a number of sensitivity checks, a
fundamental question remains unanswered: What type of growth is associated with total
BTDs? Change in net income can be divided into two components: changes in pretax
earnings and changes in tax expense4 . Holding future effective tax rates constant, tax
expense is simply a function of pretax earnings, so dividing net income changes into its
pretax and tax expense components may appear unnecessary. However, in the setting of
examining firms across extremes in book-tax differences, future effective tax rates are not
expected to remain constant, and it is an empirical question as to whether any eamings
growth/BTD relation is due to changes in pretax income or changes in tax expense. In
documenting a relation between total BTDs and growth in net income, Lev and Nissim
(2004) measure BTDs as the rank of the tax/book ratio, with taxable income calculated as
current tax expense grossed up by the current statutory tax rate. Since the tax rate is
constant across firms in a given year, this measure is really just the rank of the ratio of
current tax expense to net income (similar to an effective tax rate). This implies that what
Lev and Nissim (2004) really find is that films recording a high (low) rate of current tax
4 Changes in net income can also be due to discontinued operations and extraordinary items, which are
recorded net of tax. However, both in Lev and Nissim (2004) and in this study, income is measured
before these items, so that changes in net income can be cleanly divided into the pretax income and
tax expense components.
12
for a given amount of book income are likely to have positive (negative) growth in future
net income. The relatively high or low amount of current tax could be due to temporary
differences, permanent differences, or tax accruals. 5
Temporary differences arise when financial accounting and tax accounting record
economic events in different time periods. A common example is depreciation, in which
total depreciation expense for an asset will be the same over the life of the asset, although
in any given year it is unlikely that both the financial and tax systems will record the
same expense. These timing differences are measured by the deferred tax expense.
Because total tax expense includes both current and deferred tax expense, temporary
differences do not affect the total tax expense recorded in the financial statements.
Permanent differences are transactions recognized for financial or tax purposes, but not
both. They do impact the overall tax expense recorded in the financial statements.
Common examples include goodwill write-downs, restructuring charges, and a portion of
dividends received from other firms. Permanent differences can be measured by
removing temporary differences from total BTDs.
Lev and Nissim (2004) and Weber (2008) consider tax accruals as a third
component ofBTDs, as the behavior of this component is different than the other
differences. However, short of extensive hand collection, in empirical tests tax accruals
will be treated as either temporary or permanent differences. Because the valuation
allowance account is captured as a deferred tax, the VAA will be considered empirically
I use the expression' relatively high or low amount of current tax' to compare the firm both to its
own typical tax levels in a time series, as well as to the typical tax levels faced by firms within the
same industry.
13
(and in my hypotheses) as a temporary difference. Tax accruals such as tax contingencies

and pennanently reinvested foreign earnings (PRE) will be treated empirically as
pennanent differences. Each of these BTDs (tax accruals, pennanent differences, and
temporary differences) has different implications for future earnings changes.
Tax Accruals
Consider a finn with constant underlying earnings and a constant effective tax
rate, so that the growth in net income is zero. If the effective tax rate were instead to vary
annually through tax accruals (for example, adjustments in the tax contingency account,
due either to opportunistic management behavior or underlying economic reasons), future
changes in net income would be due to changes in tax expense, unrelated to any change
in economic perfonnance. In years with a relatively high tax rate (and expense), future
rates would tend to be lower, reducing tax expense and increasing net income, with the
opposite occurring in years with a relatively low tax rate. In the case of book-tax
differences due to tax accruals, any future earnings growth related to the BTD will be due
to tax expense changes, rather than changes in core earnings. Given the wide variation in
effective tax rates finns experience from year to year, to the extent these fluctuations are
due to tax accruals, there will be a significant BTD/earnings growth relation arising
solely from tax expense changes. 6
As anecdotal evidence for the preceding, consider the book-tax differences and
subsequent earnings changes for two large finns in 2004: Johnson and Johnson and Coca-
6 Although Dyreng, Hanlon, and Maydew (2008) find some evidence of effective tax rate persistence,
they find considerable variation in them across finns and across time.
14
Cola (Table 1). The BTD is measured as taxable income minus book income, as a
percentage of assets. Johnson & Johnson had (for its industry and year) a relatively high
BTD of 0.009, indicating taxable income exceeded book income by 0.9% of total assets.
This was relatively high compared to the industry mean (-0.097), where book exceeded
tax by 9.7% of assets. Coke's BTD of -0.073 indicates that book income exceeded
taxable income by 7.3% of assets, which was relatively low compared to its industry
mean (-0.028), where book exceeded tax by 2.8% of assets. This contrast in BTDs
corresponds to relatively higher (lower) earnings changes for Johnson & Johnson (Coke)
in the following year, as predicted by prior literature. However, the change in earnings
was not due to changes in pretax earnings, but due to substantial reductions (increases) in
tax expense for Johnson & Johnson (Coke), as these firms were in the process of
Table 1
Example of Relation between Book-Tax Difference and Future Performance
Firm (2004)
Johnson & Johnson
Coca Cola
BTD as percent of assets

Industry mean
0.9%
-7.3%
-9.7%
-2.8%
Earnings measure
~pretax income
~tax expense
~net income
1.5%
-2.0%
3.5%
1.5%
1.5%
0.0%
Industry mean ~NI
-0.4%
0.1%
BTD is difference between taxable and book income (taxable income minus book income), as a
percentage of current year total assets. Taxable income is computed as current tax expense, grossed up
by statutory tax rate t. and taken net of tax by multiplying by (1 - t). Change in pretax income, tax
expense, and net income is calculated as next year change in that measure as a percentage of current
year total assets.
15
repatriating foreign earnings under the American Jobs Creation Act of2004. While only a
simple example, this suggests that much of the BTD/earnings growth relation may be due
in part to variation in tax rates caused by tax accruals.
Table 2 provides additional evidence on the mean reversion of ETRs. I form three
portfolios of firms based on a ranking of their BTD in year t, with the highest quintile
forming the 'High' portfolio, the lowest quintile forming the 'Low' portfolio, and the
other quintiles forming the 'Middle' portfolio. I compute the mean effective tax rate
(ETR) for firms in each pOlifolio in the year of their ranking, as well as in each of the five
years preceding and after the ranking. Firm-years with the highest ranking of the BTD
averaged an ETR of 45% in the year of that ranking, with ETRs dropping to 40% the
following year, and retreating to around 36% within five years, while firms with the
lowest ranking ofBTD averaged an ETR of only 20% in the year of the ranking,
climbing to 25% within five years. This suggests that firm-years with high or
Table 2
Means of Effective Tax Rates (ETR)
Across Ranks of BTD for Five Years Prior to and After BTD Ranking
R BTD
High
Medium
Low
t-5
0.39
0.36
0.23
t-4
0.39
0.36
0.22
t-3
t-2
t -1
0.39
0.35
0.21
0.40
0.35
0.21
0.40
0.35
0.20
t+l
0.45
0.36
0.21
0.40
0.36
0.23
t+2
0.38
0.35
0.24
t+3
t+4
t+5
0.38
0.35
0.24
0.37
0.34
0.24
0.36
0.34
0.25
Time t captures mean ETR for all firm years with high relative BTD (upper quintile), low relative BTD
(lowest quintile), or medium BTD (middle three quintiles), where R_BTD is the quintile ranking (within
an industry and year) of a firm's BTD scaled by assets. BTD is the difference between taxable and book
income (taxable income - book income). ETR is measured by scaling total tax expense (current +
deferred) by taxable income (pretax income less grossed up deferred taxes). ETR measure is trimmed by
deleteing any observation where the ETR measure is greater than 1 or less than -1.
16
low rankings ofBTDs quickly experienced large changes in their ETR, which in tum
could significantly affect their future after-tax net income. To the extent these swings in
ETRs are caused by tax accruals, the positive relation between tax accruals and future
changes in net income will be largely due to the negative relation with future changes in
tax expense.
Permanent D~tlerences
The relation between future eamings changes and permanent differences is less
clear. Consider a firm with otherwise constant net income, but experiences a one-time
charge that results in a permanent book-tax difference (for example, write-off of
goodwill). This BTD will result in a relatively high effective tax rate for the level of net
income (net income was reduced by the charge, with no corresponding reduction in
taxable income). Although all other economic performance is equal, future net income
will increase due to the future absence of the one-time charge, this time related to
changes in pretax income, not changes in tax expense. 7 Thus, some pennanent
differences simply capture transitory economic events that naturally affect future eamings
changes by failing to repeat. Consider now a firm with otherwise constant net income, but
with pennanent differences that are not transitory (such as dividends regularly received
from another firm, which are largely excluded from taxable income). As long as the
permanent difference is constant, there is no reason to expect a change in future net
The argument could be made that future economic perfonnance will not be equal. The write-down of
goodwill may well predict future declines in economic perfonnance. Ultimately the impact of
transitory pennanent differences on future pretax earnings changes is an empirical question.
7
17
income from pretax changes or from tax expense variation. In this case, permanent BTDs
have no relation to future earnings growth.
The previous examples demonstrate that BTDs generated by tax accruals lead to
tax expense changes, while permanent differences may be related to future changes in
pretax earnings in either direction, or they may have no relation at all to pretax earnings
changes. The third type ofBTD, temporary differences, is the most easily separated from
the other two (as it is reported as deferred tax), and will be discussed below. Controlling
for temporary differences, I propose that the relation between the non-temporary
components of BTDs and future earnings changes is due primarily to future changes in
tax expense. My first hypothesis is:
HI:
There is a negative relation between the non-temporary components of
BTDs (tax accruals and permanent differences) and future tax expense.
Temporary differences (identified by the deferred tax expense) capture income or
expense items that are recognized at different times for book and tax purposes. These
timing differences can be useful in identifying managerial discretion in accounting
decisions. For example, managers exercise judgment with respect to depreciation periods
and methods, revenue recognition, and recording reserve allowances such as bad debt,
warranties, and contingencies. To the extent that management exercises its discretion in
these matters to inflate (reduce) income for financial reporting purposes, the reversing
nature of accruals dictates that any resulting temporary BTDs will be associated with
18
future declines (increases) in pretax earnings. This is true for at least two reasons. First,
all else equal, the earnings in a year following one with inflated earnings will lower by
definition. Second, accruals reverse, so to the extent that temporary BTDs capture
accrual-based earnings management, the reversal of these managed accruals in a
subsequent year will cause future income to be lower.
Management discretion in financial reporting also serves as a signal of
management's private information about future firm performance. Consider a firm in
distress that provides for a large valuation allowance against its deferred tax assets. 8 This
creates a temporary difference reflected in a deferred tax expense, and this temporary
difference is informative about future declines in economic performance. 9 Hence,
whether temporary differences reflect the discretion in accruals used to manipulate
earnings, or rather reveals management's private information through the discretion in
accruals, there is reason to expect a negative relation between future earnings and
temporary differences. Hanlon (2005) finds that the pretax earnings of firms with large
temporary BTDs are less persistent. In contrast, Lev and Nissim (2004) find no relation
between net income growth and temporary differences. I predict that pretax earnings
8 As discussed previously, the valuation allowance account is a tax accrual, but because it is measured
by deferred taxes, it is treated as a temporary difference, both in the hypothesis development and in
the empirical tests.
9 There are cases where the creation of a deferred tax asset would predict future declines in economic
performance. For example, consider a finn constant in size, but with changes in its allowance for
doubtful accounts or warranty liabilities. These changes will generate deferred tax assets that are also
informative about future firm prospects, as it contains infonnation about management's assessment of
future collectability of receivables or quality of goods sold. Ultimately it is an empirical question
whether deferred tax expense, absent earnings management, predicts higher or lower growth.
19
growth for firms with large temporary differences will be lower. My second hypothesis
IS:
H2:
Large temporary BTDs (deferred taxes) are negatively related to growth in

pretax earnings.
Finding empirical support for H2 suggests that temporary differences are related
to future changes in pretax income, but provides no evidence as to the nature of the
causality of this relation. As discussed above, manipulated earnings numbers can lead to
temporary BTDs that will be related to lower future earnings growth. However,
underlying economic events and conditions can also result in temporary BTDs that are
related to future earnings growth. In the example noted above, a change in the valuation
allowance account results in a change in deferred taxes that is informative about future
finn prospects. Similarly, consider a firm with significant unearned revenue that is
recognized as income for tax purposes. This will result in a deferred tax asset (a negative
deferred tax expense), which will be related to future book revenue (when the previously
unearned revenue is recognized). Thus, I predict that even in the absence of earnings
management, temporary BTDs, as measured by the deferred tax expense, will be
negatively related to growth in pretax earnings, and that the relation between temporary
differences and future pretax earnings growth will be stronger in finn-years suspected of
earnings management. Addressing the causality of the relation predicted in H2, my third
set of hypotheses are:
20
H3a:
In the absence of earnings management, temporary differences are
negatively related to pretax earnings growth.
H3b: The negative relation between temporary differences and pretax earnings
growth is stronger in the presence of earnings management.
HI predicts a negative relation between permanent differences and future changes
in tax expense. However, finding empirical support for HI provides no evidence as to the
nature of the causality of this relation. As discussed in HI, the non-temporary
components of total BTDs are composed of both tax accruals (not including the valuation
allowance account) and permanent differences. While it is unclear how permanent
differences are used in earnings management, prior research has provided evidence that
tax accruals have been used to manage earnings. The manipulation of these accruals do
not affect pretax earnings, but only affects tax expense, so I expect that firms suspected
of earnings management will have a stronger negative relation between non-temporary
BTDs and future tax expense changes. However, even in the absence of earnings
management, I expect a negative relation between the non-temporary differences and
future tax expense changes as the tax expense mean reverts towards the statutory tax rate.
Consider an international firm with constant earnings. For economic reasons (investment
opportunities domestically and abroad), the firm in year one repatriates its foreign
earnings, resulting in a relatively high amount of cunent tax expense compared to its
book income (and thus a higher BTD). In the following year the firm designates its
foreign earnings as permanently reinvested (PRE), escaping the repatriation tax expense,
and hence its net income has increased due to a tax expense decrease, all related to the
21
relatively high BTD from year one. This relation occurs in the absence of earnings
management objectives. However, Krull (2004) demonstrates that the PRE designation is
used to manage earnings, so in the presence of earnings management, I expect the
relation between tax accruals and future changes in tax expense to be even stronger.
Addressing the causality of the relation predicted in HI, my final set of hypotheses are:
H4a:
In the absence of earnings management there is a negative relation
between non-temporary BTDs and future tax expense changes.

H4b: The negative relation between non-temporary BTDs and future tax
expense changes is stronger in the presence of earnings management.
22
CHAPTER IV
RESEARCH DESIGN
To test the relation between the components ofBTDs and earnings changes, I first
measure the temporary and permanent components of BTDs with the following
procedure introduced by Weber (2008): First, I estimate TAX DIFF (total BTDs) as the
difference between taxable income and net income, scaled by total assets,
TAX DIFF = (taxable income - net income) / average assets
(1)
where net income is measured as income before extraordinary items (Compustat #18) and
average assets is the mean total assets (Compustat #6) over the previous two years.
Taxable income is estimated by grossing up current tax expense,
Taxable income = current tax expense / t
* (l-t) / average assets
(2)
where the current portion of the income tax expense is grossed up by t, the top statutory
corporate federal tax rate. 10 Taxable Income is multiplied by (1 - t) to make it
comparable to Net Income, which is measured after tax.! I I then estimate TEMP, the
temporary component of total BTDs, by grossing up the negative of deferred taxes,
]0 The top statutory corporate tax rate was 48% in 1973-1978,46% in 1979-1986,40% in 1987,34%
in 1988-1992, and 35% in 1993-2006.
11 The estimate of taxable income contains measurement error from several sources, such as the use of
the top statutory tax rate in a progressive system or to represent foreign tax rates, the misalignment of
tax expense and benefits for stock options, and tax credits (See Manzon and Plesko (2002), Mills,
Newberry, and Trautman (2002), McGill and Outslay (2002), Hanlon (2003) Mills and Plesko (2003),
23
TEMP = - (deferred tax expense) / t
* (1-t) / average assets
(3)
where the negative of deferred taxes are grossed up by t, multiplied by (1 - t), and scaled
by total assets, making it comparable in measurement to Taxable Income. Because
TAX
DlFF
captures the extent to which taxable income exceeds book income, I use the
negative of deferred tax, which otherwise would capture the extent to which bo?k income
exceeds taxable income. Finally, I estimate PERM, the non-temporary component of total
BTDs (permanent differences and tax accruals), as the difference between TAX DlFF and
TEMP.
12
PERM = TAX DIFF - TEMP
(4)
This procedure captures the extent to which taxable income exceeds book income,
breaking this difference into permanent and temporary components.
and Lev and Nissim (2004) for a discussion of the measurement error in estimates of taxable income).
However, Lev and Nissim (2004) use the same estimate of taxable income in the computation of their
tax to book ratio (TAX) and find that these errors do not systematically affect the relation between the
TAX ratio and growth in net income.
12 This is a key departure from the Lev and Nissim (2004) methodology due to the difficulty in
interpreting their coefficients. Their key variables are R_TAX and R_DEF, the quintile ranks of the
taxlbook ratio and deferred taxes. Conceptually, TAX captures total BTDs, while DEF captures
temporary differences, making DEF a subset of TAX. When both are included in a regression, the
coefficient on TAX will capture the effect of permanent differences and tax accruals on growth, while
the sum of the coefficients on TAX and DEF will reflect the impact of temporary differences. Hence, a
lack of significance on DEF alone would not necessarily indicate that temporary differences do not
predict earnings growth. However, because TAX is measured as a ratio, while DEF is the temporary
component scaled by assets, DEF is no longer strictly a subset of TAX, and the interpretation of the
coefficients on these variables (or their ranks) is unclear. My method of splitting the total BTD into its
two components provides two variables that are not subsets of each other, but in sum capture the total
BTD. The coefficient on PERM now clearly captures the impact of pennanent differences and tax
accruals, while the coefficient on TEMP relates only to the impact of temporary differences.
24
I then estimate the relation between the components of BTDs and earning changes
with the following equation:
L"1NI
a + ~lPERM + ~2TEMP + ICONTROLS + E
(5)
where L"1NI is an indicator of future changes in net income. It is alternatively measured as:
next year's net income minus current net income (L"1NI 1), average net income over the
next three years minus current net income (L"1NI 3), and average net income over the next
five years minus current net income (L"1NI 5)' Net income is measured as Compustat #18
(income before extraordinary items) scaled by total assets (#6).
Because many BTDs result from accrual estimates, it could be argued that any
relation between BTDs and growth simply proxies for the effect that accruals have on
earnings growth. To examine whether permanent differences (PERM) or temporary
differences (TEMP) contain incremental infoTInation relative to accruals and cash flows, I
include two related control variables, ACC and CASH 13, estimated as total accruals and
total cash from operations, respectively, each scaled by total assets. 14 In this way, the
coefficients on PERM and TEMP should capture the information in peTInanent
differences and temporary differences incremental to each other and to accruals about
future changes in net income.
13 Cash from operations is measured as the difference between income (before extraordinary items)
and accruals, where accruals = (L'1cunent assets - L'1cash) - (L'1cunent liabilities - L'1short-tenn debt) L'1defened tax liabilities - depreciation.
14 This is similar in spirit to Lev and Nissim's (2004) CFO measure, the percentage of net income due
to cash flows. However, the CFO measure is scaled by net income, which does not allow this
cash/accrual measure to compete on an even footing with PERM and TEMP, which are scaled by total
assets.
25
Chan, Karceski, and Lakonishok (2003) and Fama and French (2000) identify
several other predictors of earnings growth. Following Lev and Nissim (2004), I add the
following control variables: current change in ROA, average change in ROA over three
and five years, dividends scaled by assets, the ratios of R&D to sales and capital
expenditures to sales, and the current earnings I price (E/P) ratio and book-to-market
(BTM) ratio. 15 Current and longer term changes in ROA control for short and long term
trends in earnings, and should be positively related to future earnings changes. The level
of dividends may reflect management's confidence in future earnings strength,
suggesting a positive relation with future earnings changes, or it may signal fewer
investment opportunities for the firm, suggesting a negative relation with future earnings.
changes. The ratios of R&D and capital expenditures to sales controls for expected sales
growth due to new investments, but it also identifies growing firms making large
investment outlays whose profitability may not improve in the short tenn, so I do not
make a prediction on these variables. The E/P and BTM ratios capture market
expectations of future growth. Each has stock price in the denominator, so that higher
stock prices (and higher market expectations for earnings growth) will result in lower
values for these ratios. Thus I expect an inverse relation between these ratios and future
changes in earnings growth.
15 Lev and Nissim (2004) also include current return on assets, controlling for the tendency of
profitability to mean revert. However, I have already captured ROA with the ACC and CASH
variables, whose numerators add to total net income, and whose denominator is total assets.
26
To examine the statistical significance of the relation between the BTD measures
and the two components of earnings changes, I estimate the following two additional
equations:
L'.PRETAX = a + ~lPERM + ~2TEMP + ~CONTROLS + E
(6)
L'.TAXEXP = a + ~lPERM + ~2TEMP + ~CONTROLS + E
(7)
where L'.PRETAX captures future changes in pretax earnings, and L'. TAXEXP captures
future changes in tax expense. Following Hanlon (2005), PRETAX is measured as pretax
income less minority interest scaled by average total assets. TAXEXP is measured as
total income tax expense scaled by total assets. As with the variation on L'.NI, L'.PRETAX
is alternatively measured as: next year's pretax earnings minus current pretax earnings
(L'.PRETAX 1), average pretax earnings over next three years minus current pretax
earnings (L'.PRETAX 3), and average pretax earnings over next five years minus current
pretax earnings (L'.PRETAX 5)' The three estimates ofTAXEXP are measured in the same
way. To the extent permanent and temporary differences are positively related to changes
in pretax earnings, the coefficients on PERM and TEMP will be positive in the estimates
of equation (6). H2 predicts a positive coefficient on TEMP, as temporary differences
(the extent to which taxable income is greater than book income due to timing
differences) are predicted to be positively related to future pretax earnings changes. 16 If
16 Note that H2 predicts a negative relation between deferred tax and future pretax earnings. To be
consistent with the measurement of PERM and TAX DlFF, which measure the extent to which taxable
27
the BTD measures are negatively related to changes in tax expense, then the coefficients
will be negative when estimating equation (7). HI predicts a negative coefficient on
PERM when estimating equation (7), as the tax accruals component of PERM is expected
to be related to declines in tax expense.
If a component ofBTDs predicts eamings growth, it could reflect the influence
of current eamings management activities on future growth in eamings, or it may simply
reflect underlying economic events that generate various levels of BTDs and are related
to eamings growth. To examine this issue, I identify finn-year observations where
eamings management is suspected and create a dummy indicator (EM) equal to one for
these finn years, and zero for finns that have no evidence of eamings management.
Similar to Phillips, Pincus, and Rego (2003), I use four different proxies for eamings
management: avoiding an eamings decline, avoiding an eamings loss, meeting analysts'
forecasts, and high discretionary accruals. Prior work such as Burgstahler and Dichev
(1997) indicate that an abnonnally high percentage of finns have small eamings increases
or slightly positive eamings, suggesting that many finns in these categories are managing
their eamings upwards to avoid an eamings decrease or a loss. To identify eamings
management based on eamings changes, EMl =1 if the change in net income (Compustat
#172) scaled by previous year's beginning market value of equity (#25 x #199) is
~O
and
<0.02, and EMI =0 otherwise. To identify eamings management based on avoiding a loss,
I compare finns with zero or slightly positive scaled eamings with those that easily
income exceeds book income, TEMP is the negative of deferred taxes scaled by assets, so the
expected coefficient on TEMP is positive.
28
attained positive earnings. Specifically, EM2=1 ifnet income scaled by beginning of year
market value is ~O and <0.02, and 0 otherwise.
Degeorge, Patel, and Zeckhauser (1999) find evidence that suggests earnings
management among firms that narrowly beat analysts' forecasts. To identify earnings
management with analysts' forecasts, I identify firms that meet or narrowly beat their
forecast. Specifically, EM3=1 if the earnings surprise (actual IBES earnings per shareconsensus forecast) is ~O and less than 0.02, and EM3=0 otherwise. I?
Discretionary accruals have also been proposed as a means of identifying earnings
management. I follow the modified Jones model proposed by Dechow, Sloan, and
Sweeney (1995) to identify discretionary accruals. I estimate the following equation:
Accruals = a + ~l(l/A) + ~2C~REV-~REC) + ~3(PPE) + E
where accruals are as measured earlier, A is total assets,

(Compustat #12),
~REC
~REV
(8)
is change in revenue
is change in receivables (#2), and PPE is property, plant, and
equipment (#7), each scaled by total assets (#6). Equation (4) is estimated crosssectionally each year for each two-digit SIC code. The residuals from this equation are
categorized as discretionary accruals, and I then rank the residuals into quintiles. I set the
earnings management indicator EM4=1 for firms with the highest quintile rank of
discretionary accruals, and 0 otherwise.
17 In sensitivity tests, I use various other thresholds for 'small' earnings changes, 'slightly' positive
earnings, and 'narrowly' beating analysts' forecasts. Results are substantially unchanged.
29
With these proxies for earnings management, I then re-estimate equations (1), (2),
and (3), including the EM dummy alone and interacted with the rank variables for nontemporary BTDs (PERM), temporary BTDs (TEMP), and all control variables:
i'1NI
a + ~lPERM + P2TEMP + P3EM + P4EM*PERM +

~5EM*TEMP
+ LCONTROLS + EM*LCONTROLS +
(9)
i'1PRETAX = a + P1PERM + ~2TEMP + ~3EM + ~4EM*PERM +

~5EM*TEMP
i'1TAXEXP
(10)
a + ~IPERM + ~2TEMP + ~3EM + ~4EM*PERM +
~5EM*TEMP
(11 )
where i'1NI, i'1PRETAX, and i'1TAXEXP are alternatively measured over one, three, and
five year periods as before, and EM is alternatively identified as suspected earnings
management finn-years based on earning changes, earnings levels, analysts' forecasts, or
discretionary accruals. I continue to control for growth related variables as discussed
earlier, and each independent variable is interacted with the EM dummy. If finns that
manage earnings are expected to have less future growth,
~3
will be negative. As
predicted by H3a and H4a, if the relation between BTDs and earnings changes is not only
an artifact of earnings management, but has underlying economic causes, the coefficient
on TEMP in equation (10) will be positive, and the coefficient on PERM in equation (11)
will be negative. As predicted by H3b and H4b, if earnings management intensifies the
magnitude of the BTD/growth relationship, the coefficient on EM*TEMP in equation
(10) will be positive, and the coefficient on EM *PERM in equation (11) will be negative.
30
CHAPTER V
SAMPLE AND SUMMARY STATISTICS
Sample Selection
I draw my sample from the annual Compustat files for years 1973-2006. I restrict
my sample to firms that are incorporated in the U.S. (Compustat FINC=O), are not a
financial, utility, or flow-through entity (SIC codes 4000s and 6000s), and have a
December year-end (FYR=12). These restrictions are necessary as foreign-incorporated
firms face different accounting and tax rules, utilities and financial institutions face
different regulatory and reporting rules, and flow-through entities do not pay tax.
Requiring a common fiscal year controls for temporal fluctuations in the economy.
Because domestic and foreign components of current and deferred tax are not widely
available on Compustat before 1973, I begin my sample selection in that year.
To perfonn my tests, my initial sample includes only observations that have the
following data: Total assets (Compustat #6), total income tax (#16), income before
extraordinary items (#18), number of shares outstanding (#25), deferred taxes (#50),
common equity (#60), and price per share (#199). Due to the difficulty of interpreting
BTDs for firms with negative income and taxes, I include only observations with positive
net income and tax expense. 18 This selection procedure results in a base sample of 49,956
finn-year observations, representing 6,837 different finns over the 34-year period 19732006. To mitigate the influence of extreme observations, I delete from each analysis
]8
In sensitivity tests I relax this restriction, with no change in results.
31
observations in which any continuous variable lies beyond the highest and lowest 0.5% of
the distribution for that variable. 19
Summary Statistics
Table 3 presents a comparison of means of the independent variables (including

the control variables for growth) across high and low quinti1e ranks of the PERM and
TEMP variables. The means are significantly different for all of these variables across
different quinti1es. This highlights how these firms are fundamentally different from one
another depending on their level of permanent or temporary BTDs and the importance of
controlling for these growth proxies when determining the impact ofBTDs on future
earnings growth. Table 4 presents a correlation matrix for these variables. As expected,
there is a high degree of correlation between changes in net income, changes in pretax
income, and changes in tax expense. As predicted by H2, there is a positive correlation
between temporary BTDs and changes in pretax income. In contrast to expectations, there
is a positive correlation between permanent BTDs and changes in tax expense. However,
the high degree of correlation between the dependent and independent variables makes
inference difficult in a univariate setting, highlighting the need for multivariate tests.
19 An exception to this is R&D expenditures. Because I set this variable to zero if missing in
Compustat, over half of the observations have zero R&D, and selecting a lower 0.5% to delete is not
feasible. I do exclude observations with R&D in the upper 0.5% of the distribution for regressions
including this variable.
32
Table 3
Comparison of Variable Means Across Extremes in BTDs
Variable
PERM
TEMP
ACC
CASH
HROA
HROA3
HROA5
DIY
RND
CAPX
EP
BTM
Low
(0.010)
(0.008)
(0.035)
0.105
0.009
0.015
0.018
0.022
0.012
0.075
0.101
0.857
PERM
High
0.042
(0.002)
(0.029)
0.131
0.020
0.031
0.037
0.031
0.021
0.098
0.110
0.732
Prob (t)
<0.001
<0.001
0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
Low
TEMP
High
0.009
(0.027)
(0.042)
0.125
0.016
0.025
0.030
0.022
0.014
0.112
0.108
0.760
0.016
0.013
(0.032)
0.109
0.009
0.015
0.018
0.025
0.018
0.072
0.099
0.868
Prob (t)
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.00 1
<0.001
The categories 'Low' and 'High' refer to lowest and highest quintiles of PERM and TEMP, which are
ranked by year and SIC code (2 digit). PERM and TEMP are the permanent and temporary components
of total BTDs, and are measured as the extent to which taxable income exceeds book income. Thus, a
positive value of PERM (TEMP) reflects permanent (temporary) differences that cause taxable income
to exceed book income, while a negative value reflects differences that cause book income to exceed
taxable income. For each quintile of observations, means are presented for each independent variable.
Prob (t) represents probabil ity that means are equal.
Table 4
Correlation Matrix
t.NI
llPT
0.834
<0.001
llTAX
0.729
0.012
(0.073)
0.114
0.025
<0.001
PERM
TEMP
t.PT
t.TAX
PERM
TEMP
0.027
<0.001
0.113
<0.001
ACC
(0.038)
<0.001
(0.118)
0.092
(0.042)
0.043
CASH
0.033
<0.001
(0.028)
0.006
(0.015)
0.014
(0.029)
0.027
0.026
<0.001
(0.013)
<0.001
0.006
1.000
(0.095)
<0.001
(0.176)
0.111
0.046
<0.001
HROA
HROA3
HROA5
DIV
RND
CAP
EP
BTM
(0.068)
<0.001
0.009
<0.001
0.231
(0:210)
0.013
<0.001
(0.013)
<0.001
0.052
<0.001
(0.003 )
<0.001
<0.001
0.274
0.051
0.316
<0.001
0.271
0.012
<0.001
(0.107)
0.659
(0.071 )
1.000
Pearson correlations presented, followed by prob (t).
(0.053 )
<0.001
(0.093 )
<0.001
(0.114)
<0.001
(0.133)
<0.001
0.076
0.466
0.110
0.133
(0.222)
<0.001
(0.071 )
<0.001
0.034
0.819
0.203
0.095
0.101
0.365
<0.001
0.578
<0.001
0.117
15
0.698
<0.001
0.919
<0.001
<0.001
0.087
<0.001
0.121
0.012
<0.001 <0.001
(0.051 )
<0.001
0.033
<0.001
0.014
<0.001
0.111
<0.001
0.231
<0.001
(0.348)
<0.001
(0.204)
<0.001
0.463
<0.001
0.089
0.106
(0.020)
0.366
(0.130)
<0.001
(0.259)
<0.001
W
W
34
CHAPTER VI
EMPIRICAL RESULTS
The Association between Book-Tax D~frerences and Earnings Changes

Table 5 presents results from cross-sectional regressions of equations (5), (6), and
(7). To control for differences across time and across industries, year and industry (two
digit SIC code) fixed effects are included in all tests. The first three columns present
results from using net income as the measure of earnings growth. Similar to the findings
of Lev and Nissim (2004), I find that PERM is positively and strongly related to
subsequent growth in net income over one, three, and five year periods. In contrast to the
finding of Lev and Nissim (2004), TEMP is also positively related to changes in net
income, although the coefficients on TEMP are significantly smaller than those on
PERM. 20 The following three columns of Table 5 present evidence on the association
between pretax earnings growth and the pennanent and temporary components ofBTDs.
Consistent with H2, TEMP is positively and strongly related to subsequent changes in
pretax earnings over one, three, and five year periods, suggesting that the temporary
component ofBTDs, deferred tax, is negatively related to future changes in pretax
earnings. 21 Interestingly, PERM is only weakly related to future pretax earnings change
A test of the difference in coefficients on PERM and TEMP for earnings changes over 1, 3, and 5
years report F-statistics of 2.53, 21.90, and 21.25 (p=O.ll, <0.001, and <0.001).
20
21 Because TEMP is measured as the negative of deferred tax expense, a positive coefficient on TEMP
implies a negative relation between deferred tax and future pretax earnings changes.
Table 5
Regressions of Future Earnings Changes on BTD Measures and Other Indicators of Growth
Sign
?
Intercept
PERM
TEMP
ACC
Change in Net income

3 years
5 years
1 year
0.05***
0.02***
0.04***
(5.12)
(3.54)
(2.93)
0.29***
0.18***
0.32***
(11.86)
(9.94)
(9.72)
0.13***
0.11 ***
0.14***
(4.29)
(6.04)
(2.86)
-0.02
(-1.29)
0.03*
(1.74)
0.06***
(2.70)
0.12***
(5.70)
0.00
(0.19)
-0.04
(-1.19)
0.02
(0.74)
-0.09*
(-1.88)
-0.11 ***
(-2.38)
0.13***
(4.27)
0.18***
(6.15)
0.04
(1.44)
-0.17***
(-2.68)
-0.08
(-1.26)
0.03
(0.96)
-0.04*
(-1.68)
0.00
(-0.33)
0.06
(1.46)
0.02
(0.67)
0.00
(0.32)
CASH
flROA
flROA3
flROA5
0.06
(1.60)
DIV
RND
0.01
(0.43)
-0.06***
(-3.45)
CAP
-0.01
(-1.34)
Sign
?
?
+
+
+
Change in Pretax Income

5 years
1 year
3 years
0.03***
0.05***
0.04***
(2.46)
(3.49)
(4.25)
0.06**
-0.06
0.02
(-1.64)
(2.00)
(0.65)
-0.34***
(-12.85)
-0.24***
(-9.31)
0.02
(0.73)
-0.07
(-1.18)
-0.06
(-1.14)
0.14***
(3.75)
-0.07***
(-2.48)
-0.04***
(-5.19)
Sign
?
I
+
0.15***
(11.60)
0.13***
(7.67)
0.11 ***
(5.19)
0.01
(0.55)
0.02***
(2.53)
0.02***
(2.54)
0.03*
(1.68)
-0.03*
(-1.68)
-0.04***
(-2.89)
-0.02
(-1.46)
0.00
(0.35)
0.03***
(2.42)
0.01
(0.43)
0.02
(1.42)
-0.51 ***
(-16.87)
-0.41***
(-14.07)
-0.64***
(-18.63)
-0.55***
(-16.60)
0.03
(1.04)
-0.02
(-0.27)
-0.28***
(-4.23)
0.30***
(7.19)
0.03
(0.74)
-0.09
(-1.19)
-0.28***
(-3.67)
0.41***
(8.71)
0.02
(0.45)
-0.01
(-1.30)
-0.04
(-1.03)
-0.02**
(-2.04)
Change in tax expense

1 year
3 years
5 years
0.01 ***
0.02***
0.02***
(2.52)
(4.77)
(3.22)
+
?
-0.05***
(-4.60)
0.00
(-0.43)
-0.03
(-1.13)
0.01
(0.30)
-0.02
(-0.99)
-0.04***
(-2.87)
0.00
(0.38)
0.05***
(2.87)
-0.1 0***
(-3.01)
0.05
(1.34)
-0.01
(-0.49)
-0.03*
(-1.85)
0.00
(1.15)
t.N
VI
Table 5 (continued)
Change in Pretax [ncome

Sign
EP
BTM
Observations
R2
1 year
-0.17***
(-14.92)
-0.01 ***
(-7.46)
12,459
0.119
3 years
-0.20***
(-13.74)
-0.01 ***
(-5.42)
10,592
0.157
5 years
-0.24***
(-12.66)
-0.01 ***
(-2.53 )
9,039
0.149
Sign
-
1 year
-0.17***
(-9.53)
-0.02***
(-7.75)
9,152
0.157
3 years
-0.16***
(-8.11)
-0.02***
(-6.76)
7,159
0.257
5 years
-0.13***
(-5.77)
Sign
-0.02***
(-5.86)
5,642
0.330
1 year
-0.10***
(-16.35)
3 years
-0.15***
(-18.31)
5 years
-0.17***
(-16.53)
0.00***
(-3.04)
12,438
0.104
0.00
(-0.47)
10,575
0.126
0.00
(1.13)
9,028
0.141
One year changes in net income, pretax income, and tax expense are calculated as subsequent year measure minus current, scaled by current total assets.
Three and five year changes are computed as average of measure in subsequent three or five years less current, scaled by current total assets. PERM and
TEMP are the two components ofTAXDIFF, the total book-tax difference, computed as taxable income (current tax expense grossed up by the statutory
tax rate t, net of tax (l-t)) minus net income before extraordinary items, all scaled by total assets. TEMP is the negative of deferred tax expense, grossed
up by t , net of tax (1-t), and scaled by total assets. PERM = TAXDIFF - TEMP. Thus PERM and TEMP capture the permanent and temporary
components of total BTDs, the extent to which taxable income exceeds book income. ACC is accruals scaled by total assets, computed as (lcurrent assets
less ~cash) - (~liabilities - ~ST debt) - ~deferred taxes - depreciation. CASH is cash flow from operations scaled by total assets, computed as net income
less accruals. ~ROA is current change ofreturn on assets, HROA3 and HROA5 are average change in ROA over three and five years. DIV is current
dividends scaled by assets. RND and CAPX are R&D expenditures and capital expenditures, each scaled by sales. EP is the earnings to price ratio. BTM is
the book to market ratio. Regressions include year and industry (two digit SIC code) fixed effects. *, **, *** represent statistical significance at the 10%,
5%, and I% levels (two-tailed test). Sample includes all Compustat firms with a December year end and all necessary variables from 1973 - 2006,
excluding banks and utilities, flow-through entities, and firms with foreign incorporation.
VJ
0\
37
in the following year, and there is no evidence of a relation over three or five years.
The final three columns of Table 5 present evidence on the association between
changes in tax expense and the permanent and temporary components of BTDs.
Consistent with HI, PERM is negatively and strongly related to changes in tax expense
over one, three, and five year periods. This result suggests that any relation between
permanent BTDs and future earnings changes are due to changes in future tax expense,
not changes in the underlying (pretax) earnings of the firm. The final three columns of
Table 5 also reveal a positive and significant relationship between TEMP and tax expense
changes. This is not unexpected, as TEMP is positively associated with pretax earnings,
and increases in pretax earnings should lead to increases in tax expense, ceteris paribus.
Impact ofEarnings Management on Association between Book-Tax D~fferences and

Earnings Growth
Tables 6-9 examine the impact of earnings management on the documented
association between BTDs and earnings changes. I use various proxies for earnings
management, including narrowly avoiding an earnings decline (Table 6), narrowly
avoiding a loss (Table 7), meeting or narrowly beating analysts' forecasts (Table 8), and a
high level of discretionary accruals (Table 9). H3a predicts that even in the absence of
earnings management there will be a positive relation between temporary BTDs and
changes in pretax earnings, i.e., the coefficient on TEMP is positive when 1'.PRETAX is
the dependent variable. H3b predicts that earnings management activity will intensify this
association, i.e., the interaction of TEMP with an earnings management dummy is
positive when 1'.PRETAX is the dependent variable. H4a predicts that even in the absence
38
of earnings management there will be a negative relation between non-temporary BTDs
and changes in tax expense (due to the impact tax accruals have on ETR swings), i.e., the
coefficient on PERM is negative when L1TAXEXP is the dependent variable. H4b
predicts that earnings management activity will intensify this association, i.e., the
coefficient on the interaction of PERM with an earnings management dummy is negative.
I begin by defining earnings management on the basis of annual earnings changes.
I define an earnings management dummy EM=1 when the change in earnings (scaled by
market value of equity) is greater than or equal to zero, but less than two cents, with
EM=O otherwise. I then include the EM dummy variable alone and interacted with each
independent variable as outlined in equations (9), (10), and (11). Table 6 presents the
results. Consistent with H3a and H4a, the positive (negative) coefficient on TEMP
(PERM) suggests that the relation documented earlier between the BTD measures and
L1PRETAX (1'1 TAXEXP) hold even in the absence of earnings management. Inconsistent
with H3b and H4b, there is no evidence of statistical significance on the interaction terms
of EM with TEMP and PERM. This suggests that the BTD/earnings change relation is
not a product of this type of earnings management.
I next examine earnings management defined as avoidance of losses. I define
EM=1 when net income scaled by market value of equity is greater or equal to zero, but
less than two cents. I include the EM dummy variable alone and interacted with each
independent variable as outlined in equations (9), (10), and (11). Table 7 presents the
results. Consistent with H3a and H4a, the positive (negative) coefficient on TEMP
(PERM) suggests that the relation documented earlier between the BTD measures and
Table 6
Examining Effect of EM on BTD Coefficients: Small Earnings Changes
Intercept
PERM
TEMP

Sign 1 year
3 years
5 years
0.02*** 0.05*** 0.04***
?
(3.25)
(5.35)
(3.74)
0.21
***
0.33***
0.36***
+
(9.97)
(11. 90)
(9.95)
0.14*** 0.15*** 0.14***
?
(5.43)
(4.34)
(3.24)
ACC
-0.04*
(-1.92)
0.02
(0.59)
CASH
0.08***
(2.98)
EM
EM*PERM
0.01
(0.43 )
-0.01***
(-2.49)
-0.12***
(-2.99)
-0.17***
(-3.32)
EM*TEMP
-0.01
(-0.26)
-0.06
(-0.87)
EM*ACC
0.05
(1.33)
0.09*
(1.72)
-0.11
(-1.24)
0.17***
(2.48)
EM*CASH
0.07*
(1.86)
0.09*
(1.86)
0.16***
(2.43)
-0.01 *
(-1.80)
Change in
Sign 1 year
0.04***
?
(3.77)
Pretax Income
3 years
5 years
0.05*** 0.04***
(4.38)
(2.59)
Sign
?

1 year
3 years
5 years
0.01 *** 0.03*** 0.02***
0.05
(1.54)
0.11 ***
(3.09)
-0.01
(-1.55)
-0.20***
(-2.82)
(11.14)
(7.31)
(5.03)
0.00
(0.21)
-0.04***
(-2.86)
-0.03
(-1.53)
-0.65***
(-16.55)
0.00
(-0.32)
-0.02
(-1.58)
-0.01**
1
-0.02
(-0.80)
-0.01
(-0.99)
0.02
(1.52)
0.00***
-0.10
-0.11
-0.07***
(-2.41)
-0.05
(-1.28)
-0.05
(-1.00)
0.00
(-0.04)
0.06**
(2.18)
0.09***
(2.52)
0.02
(0.96)
0.07***
(2.73)
0.10***
(2.69)
-0.40***
(-12.85)
-0.61 ***
(-17.03)
-0.74***
(-18.27)
-0.30***
(-9.92)
-0.51 ***
(-14.71)
-0.01**
(-2.12)
-0.11 *
0.16***
(2.70)
0.24***
(3.58)
0.18***
(3.09)
0.25***
(3.87)
0.25***
(3.26)
0.25***
(3.42)
0.00
Observations
12,459
10,592
9,039
9,152
7,159
5,642
12,438
9,028
10,575
2
0.122
0.162
0.155
0.160
0.263
0.107
R
0.336
0.129
0.146
EM is a dummy for earnings management, where EM= I if the change in earnings (scaled by market value of equity) is greater than or equal to zero,
but less than two cents, with EM=O otherwise. EM is interacted with all independent variables (many omitted for parsimony). All other variables,
methodology, and sample are as reported in Table 5.
U-l
\0
Table 7
Examining Effect of EM on BTD Coefficients: Narrowly Avoiding a Loss
Intercept

Sign 1 year
3 years
5 years
0.02*** 0.05*** 0.04***
?
(2.84)
(5.15)
(3.49)
PERM
TEMP
ACC
CASH
EM
EM*PERM
EM*TEMP

Sign 1 year
3 years
5 years
0.04*** 0.05*** 0.03***
?
(3.55)
(4.38)
(2.47)
0.18***
(9.34)
0.12***
(4.88)
0.28***
(11.50)
0.10***
(3.17)
0.32***
(9.81)
-0.01
(-0.85)
0.03**
(1.98)
0.06***
(2.69)
0.12***
(5.53)
-0.01 *
(-1.92)
0.42***
(2.72)
0.13***
(4.28)
0.18***
(6.06)
0.00
(-0.30)
0.34
(1.51 )
0.56***
(2.89)
0.00
(0.10)
0.44***
(3.49)
0.45***
(3.82)
0.06*
(1.87)
0.01
(0.28)
-0.07*
(-1.86)
0.14***
(10.52)
0.12***
(6.70)
0.09***
(4.32)
-0.64***
(-17.98)
-0.02
(-1.52)
0.01
(0.40)
-0.55***
(-16.03)
0.01
(0.64)
0.02***
(2.46)
0.00
(0.16)
0.02
(1.35)
-0.01
(-1.05)
0.00
-0.01
0.00
(0.02)
0.08**
(1.97)
-0.35***
(-12.87)
-0.25***
(-9.51)
-0.02**
(-2.11)
0.49***
-0.52***
(-16.58)
-0.42***
(-13.88)
-0.04***
(-3.65)
0.63***

Sign 1 year
3 years
5 years
0.01 *** 0.02*** 0.02***
?
(2.45)
(4.79)
(3.20)
1.1
0.65**
(2.30)
0.58***
0.20*** 0.29*** 0.31***
?
(3.94)
(3.03)
(3.41)
(2.92)
?
0.33**
0.00
0.38*
0.33*
0.45**
0.43**
?
0.03
EM*ACC
?
0.16*
0.13
(0.00)
(2.24)
(1.94)
(1.92)
(2.28)
(0.53)
(1.87)
(1.97)
(1.24)
0.38***
0.49***
0.04
0.42**
?
0.36**
0.44**
?
0.18**
EM*CASH
?
0.05
0.14
(0.41)
(2.23)
(2.66)
(2.22)
(2.15)
(2.60)
(2.11 )
(0.90)
(1.34)
12,459
10,592
9,039
9,152
7,159
5,642
12,438
10,575
Observations
9,028
2
R
0.127
0.162
0.129
0.160
0.151 I
0.261
0.333
0.107
0.143
EM is a dummy for earnings management, where EM= 1 when net income scaled by market value of equity is greater or equal to zero, but less than two
cents, with EM=O otherwise. EM is interacted with all independent variables (many omitted for parsimony). All other variables, methodology, and
sample are as reported in Table 5.
.j:::>.
41
~PRETAX (~T AXEXP)
hold even in the absence of earnings management. Inconsistent
with H4b, there is no apparent relation between the interaction of EM with PERM on
~ TAXEXP,
suggesting that this type of earnings management does not drive the negative
relation between non-temporary BTDs and tax expense changes. However, I do find
support for H3b, as there is a positive and significant relation between
~PRETAX
and the
EM*TEMP interaction. This suggests that while even in the absence of earnings
management temporary differences are associated with future changes in pretax income,
the management of earnings to avoid a loss increases the intensity of this relation.
I next define EM=1 when the earnings surprise (actual EPS, as reported by IBES,
less the consensus forecast) is greater than or equal to zero but less than two cents. Table
8 presents the results. As in the case of earnings changes, I find support for H3a and H4a.
The documented relations between the BTD components and the components of earnings
changes hold even in the absence of earnings management. Also as in the case of
earnings changes, I fail to find support for H3b and H4b, suggesting that earnings
management around analysts' forecasts is not the main contributor to the relation between
BTD and earnings changes.
Finally, I set the earnings management indicator EM=1 for firms with the highest
quintile rank of discretionary accruals, and then use this indicator alone and interacted
with all each independent variable as outlined in equations (9), (10), and (11). Results are
presented in Table 9. As in the previous tests, I find support for H3a and H4a. Controlling
for earnings management, there continues to be a positive and significant relation
between TEMP and
~PRETAX,
and a negative and significant relation between PERM
Table 8
Examining Effect EM on BTD Coefficients: Analysts' Forecasts
Intercept
Sign
?
PERM
TEMP
ACC
CASH
EM
EM*PERM

1 year
3 years
5 years
0.03*** 0.06*** 0.05***
(6.30)
(4.89)
(4.44)
0.13*** 0.11 ***
0.02
(5.25)
(3.49)
(0.58)
0.16*** 0.10***
0.03
(2.82)
(5.54)
(0.66)
-0.08***
(-3.78)
-0.04*
(-1.95)
-0.01
(-1.41)
0.03
(0.46)
-0.07**
(-2.31)
0.00
(-0.08)
-0.01 *
(-1.86)
0.18**
(2.19)
0.00
(0.03)
0.05
(1.47)
-0.01
(-1.46)
0.34***
(3.06)
0.02
(0.13)
0.27***
(2.55)
0.32***
(3.18)
Sign
?
?
?
-

1 year
3 years
5 years
0.05*** 0.07*** 0.05***
(4.15)
(5.25)
(3.54)
-0.05
-0.11 *** -0.33***
(-1.22)
(-2.45)
(-6.22)
-0.42***
(-11.84)
-0.33***
(-9.62)
-0.01
(-1.20)
-0.64***
(-16.15)
-0.54***
(-14.15)
-0.02*
(-1.79)
-0.70***
(-15.16)
-0.62***
(-13.95)
-0.02*
(-1.88)
0.09
0.15
(1
0.22
Sign
?

3 years
1 year
5 years
0.02*** 0.03*** 0.03***
(3.66)
(5.95)
(4.09)
0.14***
(8.60)
0.10***
(4.68)
-0.04***
(-3.02)
-0.02
(-1.44)
-0.01 **
-0.09*** -0.07***
(-5.79)
(-2.99)
-0.07*** -0.05**
(-4.38)
(-2.22)
-0.01 *** -0.01 ***
0.07***
(2.43)
-0.04
?
0.01
0.00
0.05
(-0.43)
(0.28)
(0.02)
(0.61)
?
0.11
-0.09
EM*ACC
?
-0.02
0.00
?
0.01
0.10**
0.19***
(1.35)
(-0.78)
(-0.16)
(0.02)
(0.21)
(2.15)
(3.06)
?
0.17**
0.14*** 0.23***
EM*CASH
?
0.10
0.03
0.09
0.13
?
0.05
(2.17)
(1.51 )
(0.33)
(0.80)
(1.36)
(3.10)
(1.00)
(3.85)
6,759
5,704
5,767
4,426
3,449
6,751
Observations
8,083
8,096
5,699
2
R
0.195 I
0.143
0.189
0.173
0.297
0.358
0.162
0.130
0.179
EM is a dummy for earnings management, defined by earnings surprises (the difference between the latest consensus forecast and actual earnings per
share). EM=1 if the earnings surprise is greater than or equal to zero, but less than two cents per share, with EM=O otherwise. EM is interacted with all
independent variables (many omitted for parsimony). All other variables, methodology, and sample are as reported in Table 5.
EM*TEMP
-0.08
(-0.90)
0.02
(0.22)
-+:>.
Table 9
Examining Effect of EM on BTD Coefficients: Discretionary Accruals
Intercept
Sign
?
PERM
TEMP
ACC
CASH
EM
EM*PERM
EM*TEMP
EM*ACC
EM*CASH
Observations
R

1 year
3 years
5 years
0.02*** 0.05*** 0.04***
(3.06)
(5.19)
(3.67)
0.15*** 0.27*** 0.29***
(7.47)
(10.17)
(8.33)
0.17*** 0.13*** 0.12***
(6.65)
(2.89)
(3.96)
-0.05***
(-2.38)
0.00
(0.23)
-0.01*
(-1.89)
0.12***
(2.42)
-0.15***
(-2.36)
0.18***
(3.97)
0.18***
(3.99)
12,412
0.124
0.04
(1.53)
0.10***
(4.28)
-0.01
(-1.50)
0.10
(1.64)
0.00
(0.06)
0.19***
(3.16)
0.17***
(2.94)
10,549
0.161
0.11 ***
(3.29)
0.15***
(4.88)
-0.01 *
(-1.88)
0.16*
(1.96)
-0.10
(-0.95)
0.20***
(2.55)
0.20***
(2.46)
9,005
0.153
Sign
?
?

1 year
3 years
5 years
0.04*** 0.05*** 0.03***
(3.51 )
(4.28)
(2.52)
0.03
0.04
-0.04
(0.97)
(1.05)
(-0.98)
-0.36*** -0.53*** -0.62***

(-12.23) (-15.65) (-16.16)
-0.26*** -0.43*** -0.55***
(-9.51)
(-13.70) (-15.26)
-0.01
0.00
0.00
(-0.58)
(-0.93)
(-0.55)
0.18**
-0.09
-0.15
1
1
Sign
?
0.16***
(11.09)
0.13***
(7.38)
0.13***
(5.63)
0.00
(0.06)
0.02*
(1.94)
0.00
(-0.61)
-0.04***
(-2.94)
-0.01
(-1.02)
-0.01 ***
-0.02
(-1.01)
0.00
(0.10)
-0.01**
-0.04
(-1.17)
0.04
(1.41)
0.04
(1.45)
12,391
0.105
-0.03
(-0.69)
0.13***
(3.91)
0.13***
(3.69)
10,532
0.129
-0.14***
(-2.39)
0.16***
(3.69)
0.15***
(3.38)
8,994
0.144
?
?
?
0.16**
(2.23)
0.18***
(2.60)
9,119
0.160
0.16*
(1.91)
0.18**
(2.19)
7,132
0.259
0.02
(0.21)
0.05
(0.60)
5,624
0.335

1 year
3 years
5 years
0.01 *** 0.02*** 0.02***
(2.55)
(4.87)
(3.31 )
?
?
EM is a dummy for earnings management, where EM=1 when observation is in the highest quintile of discretionary accruals, with EM=O otherwise.
Discretionary accruals are the residual from the following model: Accruals =a + ~1(1/A) + ~2(f,.REV-f,.REC) + ~3(PPE) + 1;;. EM is interacted with all
independent variables (many omitted for parsimony). All other variables, methodology, and sample are as reported in Table 5.
+:0-
44
and 6 TAXEXP. Inconsistent with H3b, I find no evidence of a stronger relation between
TEMP and 6PRETAX in the presence of earnings management. However, consistent
with H4b, I now find a negative and significant relation between 6 TAXEXP and the
EM*PERM interaction term for the one year and three periods. This finding suggests that
firms with large, income-increasing discretionary accruals experience an especially large
relation between non-temporary BTDs and future tax expense changes.
Overall, the evidence presented in Tables 6-9 suggests that the relation between
temporary differences and future pretax earnings changes, as well as the relation between
permanent differences and future tax expense changes, are primarily due to underlying
economic events. I find only limited evidence supporting the notion that the
BTD/earnings growth relation is due to earnings management.
45
CHAPTER VII
ROBUSTNESS AND SENSITIVITY TESTS
Reconciliation with Lev and Nissim (2004)
There are several departures in my research design from that employed by Lev and
Nissim (2004). For comparability with their results, I repeat my tests, using their
methodology and sample period. 22 There are three key independent variables in their
tests: R_TAX, R_DEF, and R_CFO, which are the quintile rank of the tax to book ratio,
deferred taxes, and cash flows trom operations. 23 The tax to book ratio, TAX, is
calculated as the ratio of taxable to book income, where both net income and taxable
income are measured as previously described. DEF is the negative of deferred tax
expense scaled by average total assets. As the ratio of tax to book income, a higher value
of TAX indicates more taxable income, and as the negative of the ratio of deferred tax to
average assets, a higher value ofDEF indicates fewer deferrals and thus greater taxable
income. CFO captures the relative amount of net income coming from cash f1ow, as
opposed to accruals, measured as the difference between net income and accruals, scaled
Among the differences between my tests and that of Lev and Nissim (2004) include a shorter
sample period (their cutoff year is 2000), an inclusion offinns with negative measures of tax expense,
computing coefficients and test statistics based on the Fama and MacBeth (1973) methodology of
presenting the average of the annual coefficients, multiplying dependent variables by 100, and running
ranked regressions instead of using the underlying values.
22
Because BTDs can vary across industries due to differences in capital intensity, R&D expenditures
(and credits), and other industry-related differences, industry-ranked values of these variables are
used, ranking finns each year within their two digit SIC code.
23
46
by net income. A higher (lower) rank of this variable indicates a greater share of earnings
coming from cash flows (accruals).
Figure 1 presents initial evidence on the relation between BTDs and future
earnings growth. For each measure ofBTD (total differences: R_TAX, and temporary
differences: R_DEF) as well as for the cash flow measure (R_CFO) I form three
portfolios, based on the relative measure of the ranked variable. A continuous line
follows the growth in earnings for firms with the highest quintile of the ratio, a dotted
line for those with intelmediate values, and a dashed line following the growth of
earnings for firms in the lowest quintile. Thus, the solid lines represent growth for firms
with high levels of taxable income in relation to book income, high levels of currently
taxable income in relation to tax-deferred income, and high levels of income from cash
flows in relation to income from accruals. Panel A presents results for growth in total net
income, as in Lev and Nissim (2004).24 Similar to the patterns in their study, I find that
firm-years with relatively high taxable income (high income from cash flows) experience
significantly greater growth in net income compared to firms with relatively lower
taxable income (more income from accruals). There does not appear to be any significant
difference in net income growth for various levels of deferred taxes, the proxy for
temporary BTDs. However, in Panel B, I find evidence that portfolios with higher
taxlbook ratios (high R_TAX) have significantly lower pretax earnings growth than those
with smaller ratios, while the portfolio of high R_DEF finns shows some evidence of
Lev and Nissim (2004) present two sets of graphs, pre and post SFAS 109 (1993). However, the
patterns are similar for both periods both in their study and in mine, so for conciseness, I present both
periods together. The same holds true for multivariate results, where results are similar in both
periods, so I again present both periods together.
24
47
Figure 1
Five-Year Growth in Net Income and its Components for Portfolios of Firms
Sorted by BTD and Cash Flow Measures
Panel A: Growth in Net Income
R TAX
1.5
R DEF
1.25
0.75
....,
'-_./
R CFO
1.5
1.25
1.5
.~
.,~.~.:,>1"""'"
0.75
0.75
0.5
0.5
./
0.5
0.25
0.25
1+1 1+2 1+3 1+4 1+5
1.25
0.25
I
1+1 1+2 1+3 1+4 1+5
1+ 1 1+2 1+3 1+4 1+5
R DEF
R CFO
Panel B: Growth in Pretax Income
R TAX
3
/
/
2.5
2.5
2.5
/
2
1.5
~
/
//
...........
........
......
0.5
1.5
0.5
1.5
.. ' ..
./
.~.~
""
"
./
./
'
,0'
0.5
1+]
1+1 1+2 1+3 1+4 1+5
./
1+2 t+3 1+4 1+5
t+l
t+2
t+3
t+4
t+5
Panel C: Growth in Income Tax Expense
R TAX
2.5
R DEF
/
R CFO
2.5
2.5
/
/
/
1.5
~
....,
1.5
.'
~
,':':':,:,:-0- __
0.5
0.5
I
1+1 1+2 1+3 1+4 1+5
1.5
0.5
t+l
1+2
1+3
1+4
1+5
1+ I 1+2 1+3 t+4 1+5
Each figure represents cumulative growth from year t to year t+j (j=l, 2, ... ,5) for three portfolios: a solid
line for firm-years in the highest quintile (industry-year) of the BTD or cash flow measure, dashed line for
those in the lowest quintile, and dotted line for all other observations. TAX is the ratio of tax to book
income, scaled by earnings, DEF is the negative of the ratio of deferred tax to average assets, and CFO is
the ratio of cash from operations to net income.
48
higher pretax growth. Panel C shows that finn-years with relatively high taxable income
(high R~TAX) experience significantly smaller increases in tax expense,
contributing to their overall growth advantage in net income. This provides preliminary
evidence that the earnings growth found by Lev and Nissim (2004) for finns with high
tax to book income is due to future changes in tax expense, and not due to growth in
underlying pretax earnings. It also provides initial evidence that temporary differences
contain information about future growth in pretax earnings.
Further insight into the relationships between TAX, DEF, and CFO and the
components of earnings changes can be gleaned from a multivariate examination. The
first three columns of Table 10 replicate the results of Lev and Nissim (2004), estimating
the following equation:
(12)
where all variables are as described above, and the control variables are the same as in
the main tests, with the exception that ACC and CASH are replaced by their aggregate,
ROA. Similar to the findings of Lev and Nissim (2004), I find that R_TAX is positively
and strongly related to subsequent growth in net income over one, three, and five year
periods. R_CFO is also positively and strongly associated with growth in net income.
Also consistent with the findings of Lev and Nissim (2004), there is no statistically
significant relation between R_DEF and growth in net income. The next three columns of
Table 10 present evidence on the association between pretax earnings growth and
R_TAX, R_DEF, and R_CFO, replacing ~NI in equation (7) with ~PRETAX. In
contrast with the findings on net income growth, but consistent with H2, R_DEF is
Table 10
Regressions of Future Earnings Changes on BTD Measures and Other Indicators of Growth:
Lev and Nissim (2004) Methodology
Intercept
R TAX
R DEF

Sign 1 year
5 years
3 years
1.30*** 2.38*** 2.63***
?
(3.44)
(4.68)
(3.07)
+ 0.24*** 0.21 *** 0.20***
(4.27)
(4.33)
(3.05)
0.04
?
0.01
0.00
(0.24)
(0.71)
(-0.02)
R CFO
ROA
LiROA
LiROA3
LiROA5
DIV
RND
CAP
0.16***
(4.05)
7.90***
(2.68)
0.63
(0.18)
2.30
(0.27)
-10.87
(-1.20)
-5.34*
(-1.72)
-6.81 *
(-1.85)
-0.31
(-0.47)
0.19*** 0.20***
(4.28)
(3.68)
14.74*** 27.37***
(4.20)
(4.95)
2.62
5.75
(0.60)
(1.02)
-15.44
-1.48
(-0.16)
(-1.48)
-13.81
-12.11
(-1.50)
(-1.04)
-4.96
-3.45
(-1.06)
(-0.48)
-2.05
3.47
(-0.53)
(0.80)
-0.45
-0.35
(-0.62)
(-0.40)

Sign 1 year
3 years
5 years
0.24
1.67**
?
0.76
(0.27)
(2.28)
(1.02)
-0.13*** -0.32***
?
0.10
(1.29)

Sign 1 year
3 years
5 years
0.75*** 1.68*** 2.21 ***
?
(2.53)
(5.11 )
(4.32)
(5.22)
0.35***
(4.90)
-6.67
(-1.34)
0.50
+
(0.08)
11.23
+
(0.97)
-41.08***
+
(-3.06)
?
3.74
(0.63)
?
-5.68
(-1.15)
-2.18**
?
(-2.30)
0.35***
(6.58)
-22.74***
(-5.53)
7.08
(1.11)
13.78
(1.31)
-59.09***
(-5.82)
19.34***
(3.21 )
3.38
(0.65)
-0.46
(-0.58)
0.34***
(4.22)
-38.21 ***
(-6.54)
6.09
(0.78)
-6.71
(-0.67)
-48.83***
(-4.52)
36.57***
(3.79)
8.70
(1.47)
-0.95
(-1.41)
+
+
+
?
?
?
0.06***
(3.83)
2.31
(1.01 )
0.99
(0.47)
4.17
(1.02)
-5.27
(-0.99)
-2.48
(-1.12)
-3.91 *
(-1.91)
0.09
(0.26)
(6.77)
(5.81)
0.06**
0.06*
(2.32)
(1.72)
6.88*** 10.21 ***
(2.62)
(2.75)
4.28*
3.19
(1.59)
(1.66)
-1.77
-9.05
(-0.30)
(-1.25)
-4.88
-0.53
(-0.86)
(-0.07)
-7.08*** -5.42*
(-3.34)
(-1.72)
-2.41
-2.20
(-1.17)
(-0.82)
-0.21
0.39
(-0.55)
(0.92)
.j:::.
\0
Table 10 (continued)
Sign
EP
BTM
Observations
2
R

5 years
1 year
3 years
-19.25*** -23.42*** -27.60***
(-7.53)
(-8.13)
(-7.29)
-0.92*** -1.03*** -0.97***
(-4.69)
(-5.26)
(-3.49)
420
497
458
0.270
0.288
0.294
Sign
-

1 year
3 years
5 years
-17.06*** -15.31 *** -7.24**
(-4.49)
(-4.37)
(-2.24)
-1.74*** -2.15*** -2.36***
(-6.78)
(-8.63)
(-6.82)
264
366
311
0.306
0.464
0.390
Sign

1 year
3 years
5 years
-9.94*** -14.14*** -17.11 ***
(-7.85)
(-13.08) (-10.22)
-0.37*** -0.37*** -0.32*
(-3.38)
(-2.98)
(-1.86)
496
457
420
0.244
0.284
0.295
Table 10 contains the same dependent variables as Table 5, but multiplied by 100 to express in percentage points (as in Lev and Nissim, 2004).
R_TAX is the quintile rank (within industry and year) of the tax/book ratio, where taxable income is as previously defined. R_DEF is the quintile
rank of the negative of deferred tax expense scaled by total assets. R_CFO is the quintile rank of cash flow from operations scaled by net income,
where cash flow from operations is as defined earlier. All other independent variables are as previously defined. All regressions include industry
fixed effects (two-digit SIC code). Coefficients, number of observations, and ~ are the mean from a series of annual regressions, and the associated
t-statistics are calculated as the ratio of the mean cross-sectional coefficient to its standard error (Fama and MacBeth, 1973). *, **, *** represent
statistical significance at the 10%., 5%" and 1% levels (two-tailed test). Sample includes all Compustat firms with a December year end and all
necessary variables from 1973 - 2000, excluding banks and utilities, flow-through entities, and firms with foreign incorporation.
Vl
51
positively associated with pretax earnings growth, while R_TAX has either no statistical
association with pretax earnings changes or has a significant negative association. The
final three columns of Table 10 examine the association between R_TAX, R_DEF, and
R CFO with changes in tax expense, replacing L1NI in equation (7) with L1TAXEXP.
Consistent with HI, R_TAX is negatively and strongly associated with changes in tax
expense. These results suggest that the positive relation Lev and Nissim (2004) find
between the tax to book ratio and earnings growth is due to changes in tax expense, not
changes in underlying pretax earnings. Interestingly, there is a positive and significant
relation between R_DEF and L1TAXEXP, indicating higher tax expense changes for firms
with large temporary BTDs. This increase in tax expense is set against the increase in
pretax income for large DEF firms, and the two together may explain why there is no
significant relation between R_DEF and changes in net income.
Firm Fixed Effects
The results presented in Table 5 are based on cross sectional regressions of a
multi-year, multi-firm panel, with both year and industry fixed effects. Because some of
the relations I predict between the BTD components and the components of earnings
changes are mechanical in nature, it could be argued that a time series regression is more
appropriate. To test this, I replace the industry fixed effects with firm fixed effects and
. rerun equations (5), (6), and (7). Table 11 presents the results. While the basic findings
are unchanged, there are some interesting differences. Most notably, the coefficients and
t-stats for the relation between PERM and L1TAXEXP are significantly more negative,
which is not surprising given the mechanical relationship predicted between these
Table 11
Firm Fixed Effects
Sign
Intercept
PERM
TEMP
ACC
CASH
.6.ROA
.6.ROA3
.6.ROA5
DIV
RND
CAP
EP
BTM
Observations
RL
Change
1 year
0.20***
(4.82)
0.20***
(8.11 )
0.16***
(6.40)
-0.26***
(-10.08)
-0.22***
(-8.73)
0.04***
(2.56)
0.01
(0.16)
0.13***
(3.20)
0.02
(0.61)
-0.06
(-1.54)
-0.01**
(-2.02)
-0.20***
(-14.77)
-0.02***
(-10.33)
12,459
0.333
in Net income
3 years
5 years
0.07*
0.02
(1. 73)
(1.35)
0.19***
0.16***
(6.67)
(4.41 )
0.09***
0.06*
(3.38)
(1.74)
-0.43*** -0.59***
(-14.79)
(-16.50)
-0.40*** -0.57***
(-13.84)
(-16.25)
0.03
0.08***
(1.51)
(3.21 )
0.05
-0.02
(1.13)
(-0.42)
0.12***
0.26***
(2.60)
(4.73)
0.27***
0.42***
(5.98)
(7.52)
-0.02
0.09
(-0.36)
0.64)
-0.02***
-0.01
(-2.45)
(-0.99)
-0.23*** -0.25***
(-13.38)
(-14.44)
-0.01***
0.00
(-0.29)
(-4.69)
9,039
10,592
0.491
0.583
Sign
?
?

1 year
3 years
5 years
0.04
0.03
0.03
(0.81 )
(0.67)
(1.43)
-0.10*** -0.27*** -0.42***
(-10.10)
Sign
+
-
+
+
+
?
?
?
-0.82***
(-19.79)
-0.74***
(-18.34)
0.05*
(1.73 )
-0.01
(-0.25)
0.24***
(3.81)
0.07
(1.12)
-0.13*
(-1.91)
-0.04***
(-4.42)
-0.26***
(-11.63)
-0.03***
(-8.21)
9,152
0.388
-1.27***
(-30.59)
-1.22***
(-30.05)
0.01
(0.37)
0.06
(0.99)
0.21 ***
(3.38)
0.26***
(4.32)
-0.16***
(-2.33)
-0.04***
(-4.23)
-0.22***
(-10.44)
-0.02***
(-5.66)
7,159
0.597
-1.51***
(-35.87)
-1.49***
(-36.23)
0.07***
(2.42)
-0.04
(-0.75)
0.32***
(4.99)
0.34***
(5.66)
-0.03
(-0.42)
-0.03***
(-3.12)
-0.21 ***
(-9.88)
-0.01 **
(-2.06)
5,642
0.725
This table is identical to Table 5 except that industry fixed effects have been replaced by finn fixed effects.
+
+
+
?
?
?

1 year
3 years
5 years
0.13***
0.00
0.01
(5.63)
(-0.10)
(1.55)
0.13***
(9.04)
-0.1 5***
(-10.30)
-0.13***
(-9.34)
0.03***
(2.67)
0.05**
(2.16)
0.02
(0.71)
0.00
(-0.07)
-0.06***
(-2.46)
-0.01*
(-1.69)
-0.12***
(-15.54)
0.00***
(-4.07)
12,438
0.286
0.05***
(2.99)
-0.31 ***
(-17.80)
-0.29***
(-17.29)
0.02*
(1.85)
0.03
(1.05)
0.07***
(2.70)
0.10***
(3.89)
-0.02
(-0.70)
-0.01 **
(-2.29)
-0.16***
(-16.94)
0.00
(0.78)
10,575
0.436
0.00
(0.10)
-0.44***
(-20.74)
-0.43***
(-20.88)
0.05***
(3.63 )
-0.01
(-0.32)
0.15***
(4.81 )
0.26***
(7.87)
0.02
(0.60)
0.00
(0.15)
-0.16***
(-14.48)
0.00***
(2.62)
9,028
0.521
VI
53
variables and the focus on within-finn changes induced by using finn fixed effects.
Interestingly, the relation between PERM and L1PRETAX is now negative and
significant. This finding is consistent with the economic expectation oflower pretax
earnings changes for finns with large, non-deductable expenses. For example, finns with
large non-deductable expenses such as goodwill write-offs or restructuring charges will
have higher values of PERM, and these expenses may predict future declines in economic
perfonnance, and thus lower future pretax earnings.
Negative Income and Tax Expense
My base sample does not include finns with negative earnings or tax expense in
the current year. This is consistent with prior literature and is due to the difficulty of
interpreting tax variables and future earnings changes for finns with losses. To examine
the sensitivity of my results to the exclusion of these finns, I relax this restriction,
increasing my sample size to 79,263 finn year observations across 9,351 finns. Results
are presented in Table 12. I continue to find support for HI and H2, namely, PERM is
negatively related to L1TAXEXP, and TEMP is positively related to L1PRETAX.
Interestingly, the relation between PERM and L1TAXEXP is much more negative,
suggesting that loss finns with large non-temporary BTDs experience much greater
swings in their future effective tax rates.
Ranked Regressions and Measurement of Cash from Operations
Prior literature on the relation between BTDs and future finn perfonnance has
tended to use rankings of the BTD variables instead oftheir underlying values. Although
I delete observations with extreme values from my sample, it is still possible that my
Table 12
Including Firms with Negative Income and Tax
Intercept
Sign
?
PERM
TEMP
ACC
CASH
L'.ROA
L'.ROA3
L'.ROA5
DIV
RND
CAP
EP
BTM

3 years
5 years
1 year
0.03***
(3.16)
0.31 ***
(17.16)
-0.36***
(-9.71)
-0.01
(-0.55)
0.05***
(2.72)
-0.03
(-1.59)
-0.12***
(-3.11)
-0.11 ***
(-2.86)
0.00
(-0.06)
-0.05***
(-2.97)
-0.02***
(-4.26)
-0.01 ***
(-4.32)
-0.02***
(-16.02)
14,735
0.154
0.05***
(4.85)
0.42***
(19.43)
-0.62***
(-13.61)
-0.03
(-1.45)
0.03
(1.18)
-0.03
(-1.34)
-0.06
(-1.36)
-0.19***
(-4.15)
0.09***
(2.61 )
0.05**
(2.23)
-0.01***
(-2.72)
-0.03***
(-8.98)
-0.02***
(-14.87)
12,470
0.216
0.04***
(2.98)
0.42***
(15.29)
-0.60***
(-10.06)
0.00
(-0.06)
0.05
(1.64)
-0.04
(-1.64)
-0.04
(-0.68)
-0.21 ***
(-3.55)
0.15***
(3.30)
0.06**
(2.01)
-0.01
(-1.03)
-0.03***
(-5.95)
-0.02***
(-10.05)
10,561
0.167
Sign
?
?
+
+
+
?
?
?

1 year
3 years
5 years
0.04***
(3.01)
0.00
-0.34***
(-13.50)
-0.24***
(-9.81)
0.02
(0.74)
-0.16***
(-3.00)
-0.13***
(-2.63)
0.09***
(2.34)
-0.09***
H.41)
-0.04***
(-7.56)
-0.03***
(-5.51)
-0.02***
(-15.35)
10,853
0.215
0.06***
(4.34)
-0.07***
-0.54***
(-18.79)
-0.44***
(-16.04)
-0.03
(-1.24)
0.00
(0.00)
-0.32***
(-5.59)
0.32***
(7.30)
-0.03
(-0.91)
-0.02***
(-3.77)
-0.06***
(-9.44)
-0.02***
(-11.87)
8,417
0.356
0.03*
(1.89)
-0.15***
-0.71 ***
(-22.26)
-0.63***
(-20.75)
-0.03
(-0.94)
-0.04
(-0.65)
-0.26***
(-4.06)
0.51 ***
(10.45)
0.02
(0.57)
-0.01**
(-2.21)
-0.05***
(-7.15)
-0.02***
(-9.47)
6,564
0.434

3 years
5 years
Sign
?
1 year
0.01**
(2.32)
0.02***
(4.39)
0.01 ***
(2.45)
0.43***
(25.56)
-0.04***
(-4.99)
-0.02***
(-3.11)
0.01
(1.28)
0.00
(-0.07)
-0.06***
(-3.73)
-0.03***
(-2.34)
-0.01
(-1.01)
0.00***
(-2.55)
-0.02***
(-11.21)
-0.01 ***
(-11.44)
14,706
0.126
0.49***
(21.38)
-0.09***
(-7.94)
-0.07***
(-6.65)
0.01
(0.63)
-0.08***
(-3.75)
0.00
(0.17)
0.02
(1.18)
-0.01
(-0.49)
0.00
(-0.41)
-0.01***
(-8.25)
-0.01 ***
(-8.63)
12,447
0.136
0.50***
(17.52)
-0.06***
(-3.91)
-0.05***
(-3.33)
0.01
(0.51)
-0.09***
(-3.13)
-0.04
(-1.51)
0.03
(1.28)
0.00
(0.14)
0.01 *
(1.82)
-0.01 ***
(-6.28)
0.00***
(-6.62)
10,540
0.138
+
+
+
?
?
?
Observations
L
R
This table is identical to Table 5 except that the sample has been expanded to include finns with negative income and negative tax expense.
Vl
..j::,..
55
findings are partially due to exceptionally large or small values of PERM and TEMP, or
to a skewed distribution ofthese variables. To test the sensitivity of my finding to this
possibility, I repeat my tests using the quintile rankings of PERM, TEMP, ACC, and
CASH, and present the results in Table 13. 25 There are no meaningful differences from
the previous tests. Another concern is the calculation of the ACC and CASH variables.
ACC are total accruals, based upon the Balance Sheet, and CASH, cash flows from
operations, is the difference between net income and ACC. As a robustness check, I use
the measure of CASH from the Statement of Cash Flows (Compustat #308), and estimate
ACC as the difference between net income and CASH. Since data on the Statement of
Cash Flows is only generally available since 1988, I rerun my main tests using a
restricted sample beginning in 1988, and compare this with tests using the variables based
on the Statement of Cash Flows. Results (untabulated) find no significant change to my
findings using the alternative measures of cash flows and accruals.
Absolute Values
Prior literature has suggested that a possible reason why BTDs predict earnings
changes is that taxable income is a more reliable indicator of firm performance, and that
the farther book income ventures away from this 'true' measure, the lower the quality of
earnings. If this is true, it suggests that the sign of the BTD does not matter, only the
magnitude ofthe difference. To explore this, I rerun my tests using the absolute values of
PERM and TEMP. Results are presented in Table 14. While there appears to be a positive
relation between PERM and changes in tax expense (which results in net income
25 As in Lev and Nissim (2004), dependent variables are multiplied by 100 to express in percentage
points, as the coefficients on ranked variables are very small for presentation purposes otherwise.
Table 13
Ranked Regressions
Intercept
Sign
?
R PERM
R TEMP
ACC
CASH
~ROA
~ROA3
~ROA5
DIV
RND
CAP
EP

1 year
3 years
5 years
2.16***
4.02***
3.41 ***
(2.86)
(4.35)
(2.91 )
0.13***
0.19***
0.23***
(4.20)
(4.97)
(4.56 )
0.15***
0.13***
0.11 **
(5.18)
(3.54)
(2.19)
Sign
?
?
-0.18***
(-3.96)
0.04
(0.74)
0.03
(0.48)
0.13*
(1.82)
0.36***
(5.76)
0.48***
(5.79)
-0.20
(-0.12)
-4.93
(-1.40)
1.29
(0.57)
-10.30**
(-2.22)
4.80
(1.54)
-0.15
(-0.07)
-7.40***
(-4.00)
-0.87**
(-1.97)
-17.67***
(-15.65)
-8.21 **
(-2.00)
3.57
( 1.20)
-19.50***
(-3.12)
-0.76
(-0.14)
9.79***
(2.69)
-0.65
(-0.20)
-0.22
(-0.30)
-24.88***
(-13.35)
3.63
(1.34)
-6.26***
(-2.63 )
-0.56
(-1.02)
-20.68***
(-14.56)
Change in
I year
4.36***
(3.75)
-0.02
(-0.40)
Pretax Income
3 years
5 years
4.83***
2.98**
(3.74)
(2.04)
-O'(l9
-0.18***
(-1.58)
-0.56***
(-7.92)
-0.21 ***
(-2.72)
-0.52***
(-6.51 )
2.69
(0.94)
-3.63
(-0.63 )
-28.34***
(-5.72)
-2.41
(-0.73)
-9.35***
(-3.08)
-3.71 ***
(-5.39)
-19.26***
(-10.90)
3.93
(1.16)
7.65
( 1.11)
-71.22***
(-11.99)
-0.96
(-0.25)
-7.37**
(-2.12)
-1.75**
(-2.31)
-19.51 ***
(-9.89)
+
?
?
?
-0.13
(-1.52)
-0.58***
(-6.39)
-0.26***
(-2.62)
Sign
?
2.88
(0.74)
5.86
(0.73)
-87.26***
(-12.66)
2.61
(0.60)
-2.98
(-0.76)
-1.36
(-1.56)
-16.26***
(-7.38)
+
?
?
'I

I year
3 years
5 years
1.21 ***
2.35***
1.94***
(4.49)
(3.04)
(9.76)
(6.86 )
(4.84)
-0.06**
(-2.22)
-0.02
(-0.76)
0.09***
(2.45)
3.10***
(2.44)
-2.46
(-0.94)
0.39
(0.17)
0.02
(0.46)
0.12***
(2.76)
4.60***
(2.81 )
-9.94***
(-2.91)
0.02
(0.80)
2.44***
(2.54)
2.74
(1.39)
-1.08
(-0.62)
-1.35
(-1.14)
-4.21 ***
(-4.06 )
-0.17
(-0.68)
-10.33***
(-16.23)
-1.07***
-1.06***
-1.02***
-0.74***
-0.61 ***
-0.25***
-0.38
(-6.26)
(-5.27)
(-8.20)
(-3.39)
(-2.71)
(-1.46)
(-3.37)
12,459
10,592
9,039
9,152
7,159
12,438
Observations
5,642
2
0.098
0.146
0.144
R
0.113
0.141
0.233
0.295
Equivilent to Table 5, with PERM, TEMP, ACC, & CASH replaced by their quintile ranks, and dependent variables multiplied by 100.
BTM
-1.07
(-0.70)
-3.02**
(-2.24)
5.68*
(1.88)
0.04
(0.02)
-2.35
(-1.36)
0.15
(0.47)
-14.79***
(-18.29)
0.48
-16.87***
(-16.43)
0.02
(0.21 )
10,575
0.121
0.18
(1.46 )
9,028
0.139
(1.21)
V>.
0\
Table 14
Using Absolute Values of BTDs
Intercept
Sign
'?
PERM
TEMP
'?
ACC
CASH
'?
6ROA
6ROA3
6ROA5
DIV
'?
RND
'?
CAP
EP
BTM
Observations
R"
Change
1 year
0.03***
(3.87)
-0.16***
(-7.55)
0.03
( 1.07)
-0.01
(-0.56)
0.04**
(2.06)
0.03
(1.41 )
0.00
(-0.01 )
-0.03
(-0.93)
-0.02
(-0.75)
-0.07***
(-3.56)
-0.01 ***
(-2.54)
-0.18***
(-15.33)
-0.01 ***
(-7.81)
12,210
0.117
in Net income
3 years
5 years
0.05***
0.04***
(3.55)
(5.26)
-0.28***
-0.23***
(-8.68)
(-7.89)
-0.02
0.00
(-0.48)
(0.05)
0.07***
0.15***
(4.92)
(3.10)
0.20***
0.13***
(6.74)
(5.90)
0.05*
0.01
(1.72)
(0.59)
-0.03
0.00
(-0.52)
(0.02)
-0.21 ***
-0.25***
(-4.32)
(-3.85)
0.01
0.03
(0.42)
(0.66)
-0.04*
0.02
(-1.81)
(0.52)
-0.01
0.00
(-1.62)
(-0.70)
-0.24***
-0.21 ***
(-12.78)
(-14.47)
-0.01 ***
-0.01 ***
(-2.40)
(-5.01)
10,392
8,895
0.145
0.142
Sign
'?
?
Change in
1 year
0.04***
(3.66)
-0.03
Pretax Income
3 years
5 years
0.05***
0.04***
(4.37)
(2.56)
-0.09***
-0.01
(-2.49)
Sign
'?
I
'?
'?
+
+
+
'?
?
?
-0.32***
(-11.01)
-0.22***
(-7.95)
0.05
( 1.62)
0.00
(0.04)
-0.15***
(-2.56)
0.11***
(2.71 )
-0.06**
(-2.03)
-0.04***
(-6.16)
-0.19***
(-10.53)
-0.02***
(-7.06)
8,940
0.150
-0.49***
(-15.54)
-0.39***
(-12.85)
0.01
(0.36)
0.01
(0.16)
-0.32***
(-4.70)
0.30***
(6.51 )
0.00
(0.06)
-0.02***
(-2.71)
-0.16***
(-7.98)
-0.01 ***
(-6.27)
7,004
0.255
-0.62***
(-17.26)
-0.53***
(-15.34)
0.05
(1.34)
-0.08
(-1.03)
-0.31 ***
(-3.94)
0.44***
(8.49)
0.06
(1.47)
-0.01
(-1.46)
-0.13***
(-5.90)
-0.01 ***
(-5.15)
5,537
0.323
This table is identical to Table 5 except that PERM and TEMP have been replaced by their absolute value.
'?
'?
+
+
+
'?
?
?

1 year
3 years
5 years
0.01 ***
0.02***
0.02***
(2.84)
(4.59)
(2.96)
0.02
(1.36)
-0.01
(-1.38)
0.00
(0.21 )
0.03***
(2.56)
0.02
(0.94)
-0.02
(-0.78)
-0.01
(-0.85)
-0.04***
(-3.83)
-0.01 **
(-2.02)
-0.11 ***
(-16.66)
0.00***
(-2.42)
12,193
0.093
0.03*
(1.68)
-0.01
(-0.88)
0.01
(0.71)
0.03**
(2.21 )
-0.01
(-0.25)
-0.03
(-1.04)
-0.01
(-0.43)
-0.03*
(-1.83)
0.00
(-0.59)
-0.15***
(-18.28)
0.00
(0.39)
10,374
0.115
0.01
(0.35)
0.03
(1.50)
0.04***
(2.48)
0.04**
(2.15)
-0.04
(-1.20)
-0.03
(-0.84)
-0.01
(-0.28)
-0.01
(-0.63)
0.00
(0.75)
-0.17***
(-15.95)
0.00
(1.56)
8,883
0.129
Vl
-.J
58
declines), there is little evidence that the magnitude of permanent differences is
associated with future changes in pretax income. There is also no evidence of a relation
between the magnitude of temporary differences and any measure of earnings changes. I
find little support for the notion that the magnitude ofBTDs is associated with future
earnings changes.
Foreign Operations
Throughout this study I find support for HI, that permanent BTDs are negatively
related to future changes in tax expense. I hypothesize this is true because tax accruals
can cause large swings in effective tax rates. A large tax accrual that can generate this
result is PRE, the designation of foreign earnings as pennanently reinvested. To test if
PRE is driving the relation between PERM and ~ TAXEXP, I designate a dummy
variable equal to one if a firm has foreign income, and zero otherwise. I then include this
variable in equations (5), (6), and (7) and interact it with each BTD measure as well as
with each control variable. Table 15 presents the results. The negative coefficient on
PERM when examining changes in tax expense suggests that even for finns without
foreign operations there is a negative relation between other tax accruals and tax expense
changes. The negative coefficient on the interaction of the foreign operations dummy
with PERM when examining changes in tax expense suggests that PRE strengthens the
PERM-~TAXEXP
relation, in hannony with HI.
Earnings Management
The definitions of earnings management in Tables 6-9 include several choices in
research design. In an attempt to discern the sensitivity of my result to design choices, I
Table 15
Examining Effect of Foreign Operations on BTD Coefficients
Intercept
PERM
TEMP

Sign 1 year
3 years
5 years
0.02*** 0.04*** 0.04***
?
(3.29)
(2.63)
(4.67)
+ 0.21*** 0.33*** 0.40***
(8.89)
(10.89)
(9.74)
0.07***
0.04
0.02
?
(1.14)
(2.48)
(0.32)
Change in
Sign I year
0.04***
?
(3.40)
?
0.08**
(2.13)
Pretax Income
3 years
5 years
0.05*** 0.03***
(4.22)
(2.48)
0.07
(1.59)
0.03
(0.54)
Sign
?
I
?

1 year
3 years
5 years
0.01**
0.02*** 0.02***
(2.09)
(4.40)
(3.12)
0.12***
(7.53)
0.11 ***
(5.04)
0.08***
(2.91)
0.10*** 0.17***
-0.33*** -0.47*** -0.61 ***
0.00
0.03***
0.02
0.02
(3.52)
(4.37)
(0.19)
(-10.12) (-12.47) (-14.42)
(2.53)
(1.01)
(0.99)
0.05*** 0.15*** 0.19***
? -0.23*** -0.38*** -0.53***
0.05*** 0.04***
CASH
?
?
0.03
(2.52)
(5.52)
(5.24)
(-7.44)
(-10.33) (-12.93)
(4.03)
(2.40)
(1.59)
0.00
+ 0.01 *** 0.01 ***
0.01
0.00
ForOps
0.00
0.00
0.00
+
+ 0.00***
(2.87)
(2.59)
(0.80)
(1.19)
(-0.37)
(0.27)
1
(-0.10)
F*PERM
-0.06
-0.16** -0.24***
+ -0.08** -0.14*** -0.22***
?
(-2.02)
(-2.95)
(-3.41)
(-0.99)
F*TEMP
?
0.15*** 0.17*** 0.19***
0.06**
0.07
?
0.05
(3.34)
(2.81 )
(2.46)
(2.30)
(1.37)
(1.63)
-0.17***
F*ACC
? -0.09*** -0.14*** -0.15***
-0.12*
-0.07*** -0.09***
?
-0.07
?
-0.05
(-2.57)
(-3.08)
(-2.41)
(-2.64)
(-1.73)
(-1.19)
(-3.51)
(-3.43)
(-1.42)
-0.12*
F*CASH
? -0.08*** -0.10***
-0.07
?
-0.03
-0.06
-0.06*** -0.08***
?
-0.03
(-2.51)
(-2.40)
(-1.17)
(-0.55)
(-1.93)
(-0.86)
(-3.45)
(-3.35)
(-1.05)
9,152
7,159
5,642
Observations
12,459
10,592
12,438
9,028
9,039
10,575
R2
0.124
0.164
0.156
0.161
0.263
0.339
0.145
0.108
0.130
ForOps is a dummy variable equal to one if foreign income (Compustat #273) is non-missing, and zero otherwise. All independent variables are
interacted with this dummy (most interaction tenus and control variables omitted for parsimony). All other variables, methodology, and sample are
as reported earlier in Table 5.
ACC
V1
'-0
60
rerun the tests reported in those tables with the following adjustments (results
untabulated): First, my initial definition of 'narrowly' achieving an earnings target is set
at 0.02. I also use 0.01 and 0.005, with results fundamentally unchanged. Additionally,
prior literature, such as Phillips, Pincus, and Rego (2003), define earnings management as
the contrast between firms that narrowly make an earnings target vs. those that narrowly
miss. As an alternative to the results in Tables 6 and 8, I define the EM dummies in this
way (beat by 0.02 or less vs. miss by .02 or less). Results are substantially unchanged,
although some of the coefficients in support ofH3a and H4a begin to lose their
significance. These results could be due to lack of power, as a significant number of
observations are lost with this narrow definition of earnings management.
61
CHAPTER VIII
CONCLUSION
I examine the relation between book-tax differences and earnings growth. While
prior literature such as Lev and Nissim (2004) find evidence of a relation between total
BTDs and earnings growth, I find evidence consistent with temporary differences
(deferred taxes) being negatively related to growth in pretax earnings, while nontemporary difference are positively related to earnings growth only because they are
negatively related to tax expense changes. Additionally, I find these results even in the
absence of situations when earnings management is suspected, suggesting that the
relation between BTDs and earnings growth is not a result solely attributable to earnings
management, but that underlying economic events and ETR fluctuations significantly
contribute to the relation. In fact, I find only weak evidence that earnings management
contributes to the relation between temporary BTDs and changes in pretax earnings or the
relation between pennanent BTDs and changes in tax expense.
The findings in this paper contribute to our understanding of how various
measures of BTDs relate to future earnings growth. It answers the call of Graham, Raedy,
and Shackelford (2008) and Hanlon and Heitzman (2009) for investigation into the
components of BTDs. It bridges the conflicting results of two seminal papers on the
relation between BTDs and future perfonnance, as Hanlon (2005) finds a relation
between temporary BTDs and earnings persistence, while Lev and Nissim (2004) find
62
temporary BTDs to be unrelated to future earnings changes. Understanding how
temporary BTDs map into future pretax earnings growth and permanent BTDs map into
future tax expense changes can be a useful distinction for researchers investigating
investor or analyst reaction to BTDs, or when attempting to use BTDs as a measure of
earnings quality.
There are several avenues available for future research. Ayers, Jiang, and
Laplante (2009) find that taxable income is more informative among firms they identify
with 'low earnings quality' and less informative among firms designated as 'tax
planners'. Future work could consider the impact of their measures on the BTD/earnings
growth relation I have documented in this paper. Additionally, many studies examining
the impact BTDs have on market participants define BTDs with the taxlbook ratio.
Because my findings suggest that the different components ofBTDs imply different
things about future economic performance, the breakdown of BTDs into its
compon~nts
may reveal how well market participants understand these differences. Finally, the results
on earnings management are somewhat surprising and subject to various limitations.
Future work using more refined measures of earnings management will perhaps yield
results more consistent with the existing literature and the hypotheses developed in this
paper.
63
BIBLIOGRAPHY
Ayers, B., J. Jiang, and S. Laplante. 2009. Taxable income as a performance measure:
The effects of tax planning and earnings quality. Forthcoming in Contemporary
Accounting Research.
Bauman, c., M. Bauman, and R. Halsey. 2001. Do firms use the deferred tax asset
valuation allowance to manage earnings? The Journal of the American Taxation
Association 23 :27-48.
Blouin, J., and 1. Tuna. 2007. Tax Contingencies: Cushioning the blow to earnings?
Working paper, University of Pennsylvania.
Burgstahler, D. and 1. Dichev, 1997. "Earnings management to avoid earnings decreases
and losses." Journal ofAccounting and Economics, 24,99-126.
Chan, 1., J. Karceski, and J. Lakonishok. 2003. The level and persistence of growth rates.
Journal ofFinance 58: 643-684.
Dechow, P., R. Sloan, and A. Sweeney. 1995. Detecting earnings management. The
Accounting Review: 70,193-225.
Degeorge, F., 1. Patel, and R. Zeckhauser. 1999. Earnings management to exceed
thresholds. The Journal ofBusiness, 72,1-33.
Dhaliwal, D., C. Gleason, and L. Mills. 2004. Last chance earnings management: Using
the tax expense to meet analysts' forecasts. Contemporary Accounting Research
21: 431-459.
Dhaliwal, D., R. Huber, H. Lee, and M. Pincus. 2008. Variability of book-tax differences,
information uncertainty, and implied cost of capital. Working paper, University of
Arizona, UC Irvine.
Dyreng, S., M. Hanlon, and E. Maydew. 2008. Long-run corporate tax avoidance. The
Accounting Review 83:61-82
Fama, E. and K. French. 1992. The cross-section of expected stock returns. Journal of
Finance 47: 427-465.
64
Fama, E. and J. MacBeth. 1973. "Risk return and equilibrium: Empirical tests." Journal
a.! Political Economy 71 :607-636.
Frank, M.M. and S. Rego. 2006. Do managers use the valuation allowance account to
manage earnings around certain earnings targets? Journal of the American
Taxation Association 28: 43-65.
Graham, J., J. Raedy, and D. Shackelford. 2008. Research in accounting for income
taxes. Working paper, Duke University and University of North Carolina.
Gupta, S., and R. Laux. 2008. Do firms use tax cushion reversals to meet earnings
targets? Working paper, Michigan State University, Arizona State University.
Hanlon, M. 2003. What can we infer about a firm's taxable income from its financial
statements? National Tax Journal 56: 831-864.
Hanlon, M. 2005. The persistence and pricing of earnings, accruals and cash flows when
firms have large book-tax differences. The Accounting Review 80:137-166.
Hanlon, M. and S. Heitzman. 2009. Tax research: real effects, earnings management, and
governance. Working paper, University of Michigan and University of Rochester.
Hanlon, M. and G. Krishnan. 2006. Do auditors use the infonnation reflected in book-tax
differences? Working paper, University of Michigan and George Mason
University.
Healy, P. 1985. The effect of bonus schemes on accounting decisions. Journal of
Accounting and Economics 7: 85-107.
Joos, P., J. Pratt, and D. Young. 2000. Book-tax differences and the value relevance of
earnings. Working paper, Massachusetts Institute of Technology, Indiana
University, and INSEAD.
Krull, L. 2004. Permanently reinvested foreign earnings, taxes, and earnmgs
management. The Accounting Review 79: 745-767.
Lev, B., and D. Nissim. 2004. Taxable income, future earnings, and equity values. The
Accounting Review 79: 1039-1074.
Manzon, G. and G. Plesko. 2002. The relation between financial and tax reporting
measures of income. The Law Review 55: 175-214.
McGill, G. and E. Outslay. 2002. Did Enron pay taxes? Using accounting information to
decipher tax status. Tax Notes (August 19): 1125-1136.
65
Miller, G., and D. Skinner. 1998. Determinants of the valuation allowance for deferred
tax assets under SFAS No. 109. The Accounting Review 73: 213-233.
Mills, L. and G. Plesko. 2003. Bridging the reporting gap: A proposal for more
informative reconciling of book and tax income. National Tax Journal 56: 865893.
Mills, L. and K. Newberry. 2001. The influence of tax and nontax costs on book-tax
reporting differences: Public and private firms. Journal ofthe American Tax
Association 23: 1-19.
Mills, L., K. Newberry, and W. Trautman, 2002. Trends in book-tax income and balance
sheet differences. Tax Notes (August 19): 1109-1124.
Phillips, J., M. Pincus, and S. Rego. 2003. Earnings management: New evidence based
on deferred tax expense. The Accounting Review 78: 491-521.
Schrand, c., and F. Wong. 2003. Earnings management using the valuation allowance for
deferred tax assets under SFAS No.1 09. Contemporary Accounting Research 20:
579-611.
Shevlin, T. 2002. Corporate tax shelters and book-tax differences. Tax Law Review
55:427-443.
Sloan, R. 1996. Do stock prices fully reflect information in accruals and cash flows about
future earnings? The Accounting Review 71 :289-315.
Tang, T. 2005. Book-tax differences, a proxy for earnings management and tax
management - Empirical evidence from China. Working paper, The Australian
National University.
Visvanathan, G. 1998. Deferred tax valuation allowances and earnings management.
Journal ofFinancial Statement Analysis 3:6-1 O.
Weber, D. 2008. Book-tax differences, analysts' forecast errors, and stock returns.
Working paper, University of Connecticut.
Weisman, J. 2002. At firms, dual profit pictures: The gap grows between what's earned
and what's taxed. The Washington Post (October 10): AOI.

Book Tax Diff and Earning Growth

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Book Tax Diff and Earning Growth

Uploaded by

Copyright:

Available Formats

BOOK-TAX DIFFERENCES AND EARNINGS GROWTH

University of Oregon Graduate School

An Abstract of the Dissertation of

for the degree of

Title: BOOK-TAX DIFFERENCES AND EARNINGS GROWTH

to future changes in tax expense, whereas the temporary component ofBTDs is

GRADUATE AND UNDERGRADUATE SCHOOLS ATTENDED:

AREAS OF SPECIAL INTEREST:

GRANTS, AWARDS AND HONORS:

II. RELATED RESEARCH.........................................................................................

III. HYPOTHESIS DEVELOPMENT.........................................................................

IV. RESEARCH DESIGN...........................................................................................

V. SAMPLE AND SUMMARY STATISTICS

VI. EMPIRICAL RESULTS

Impact of Earnings Management on Association between Book-Tax

VII. ROBUSTNESS AND SENSITIVITY TESTS.....................................................

Reconciliation with Lev and Nissim (2004)

Firm Fixed Effects

Negative Income and Tax Expense........................................................................

Ranked Regressions and Measurement of Cash from Operations

1. Example of Relation between Book-Tax Difference

2. Means of Effective Tax Rates (ETR) Across Ranks ofBTD

3. Comparison of Variable Means Across Extremes in BTDs

5. Regressions of Future Earnings Changes on BTD Measures and Other

6. Examining Effect of EM on BTD Coefficients: Small Earnings Changes............

7. Examining Effect of EM on BTD Coefficients: Narrowly Avoiding a Loss ........

8. Examining Effect of EM on BTD Coefficients: Analysts' Forecasts....................

9. Examining Effect of EM on BTD Coefficients: Discretionary Accruals

10. Regressions of Future Earnings Changes on BTD Measures and Other

11. Regressions of Future Earnings Changes on BTD Measures and Other

12. Regressions of Future Earnings Changes on BTD Measures and Other

13. Regressions of Future Earnings Changes on BTD Measures and Other

14. Regressions of Future Earnings Changes on BTD Measures and Other

15. Examining Effect of Foreign Operations on BTD Coefficients

Underscoring the uncertainty in the literature as to which BTD measure is the

(and in my hypotheses) as a temporary difference. Tax accruals such as tax contingencies

Johnson & Johnson

BTD as percent of assets

Industry mean ~NI

There is a negative relation between the non-temporary components of

Large temporary BTDs (deferred taxes) are negatively related to growth in

In the absence of earnings management, temporary differences are

negatively related to pretax earnings growth.

HI predicts a negative relation between permanent differences and future changes

In the absence of earnings management there is a negative relation

between non-temporary BTDs and future tax expense changes.

Taxable income = current tax expense / t

* (l-t) / average assets

* (1-t) / average assets

PERM = TAX DIFF - TEMP

a + ~lPERM + ~2TEMP + ICONTROLS + E

L'.PRETAX = a + ~lPERM + ~2TEMP + ~CONTROLS + E

L'.TAXEXP = a + ~lPERM + ~2TEMP + ~CONTROLS + E

Accruals = a + ~l(l/A) + ~2C~REV-~REC) + ~3(PPE) + E

where accruals are as measured earlier, A is total assets,

is change in receivables (#2), and PPE is property, plant, and

a + ~lPERM + P2TEMP + P3EM + P4EM*PERM +

i'1PRETAX = a + P1PERM + ~2TEMP + ~3EM + ~4EM*PERM +

a + ~IPERM + ~2TEMP + ~3EM + ~4EM*PERM +

In sensitivity tests I relax this restriction, with no change in results.

Table 3 presents a comparison of means of the independent variables (including

Pearson correlations presented, followed by prob (t).

The Association between Book-Tax D~frerences and Earnings Changes

Change in Net income

-0.36* -0.53* -0.62***